mozilla
/
aom
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

remove SVC 

spatial/temporal scalability are not supported in VP10 currently.
+ remove the unused vp10/encoder/skin_detection.[hc]

this also enables DatarateTestLarge for VP10 which passes with no
experiments enabled. these were removed previously when only the SVC
tests should have been:
134710a Disable tests not applicable to VP10

Change-Id: I9ee7a0dd5ad3d8cc1e8fd5f0a90260fa43da387c
This commit is contained in:
James Zern 2016-08-08 15:09:30 -07:00
Родитель b869de9856
Коммит cc73e1fcd4
21 изменённых файлов: 37 добавлений и 4510 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

1
configure поставляемый
Просмотреть файл

@ -250,7 +250,6 @@ HAVE_LIST="
unistd_h
"
EXPERIMENT_LIST="
spatial_svc
fp_mb_stats
emulate_hardware
var_tx

8
examples.mk
Просмотреть файл

@ -109,14 +109,6 @@ endif
vpxenc.GUID = 548DEC74-7A15-4B2B-AFC3-AA102E7C25C1
vpxenc.DESCRIPTION = Full featured encoder
EXAMPLES-$(CONFIG_ENCODERS) += vpx_temporal_svc_encoder.c
vpx_temporal_svc_encoder.SRCS += ivfenc.c ivfenc.h
vpx_temporal_svc_encoder.SRCS += tools_common.c tools_common.h
vpx_temporal_svc_encoder.SRCS += video_common.h
vpx_temporal_svc_encoder.SRCS += video_writer.h video_writer.c
vpx_temporal_svc_encoder.SRCS += vpx_ports/msvc.h
vpx_temporal_svc_encoder.GUID = B18C08F2-A439-4502-A78E-849BE3D60947
vpx_temporal_svc_encoder.DESCRIPTION = Temporal SVC Encoder
EXAMPLES-$(CONFIG_DECODERS) += simple_decoder.c
simple_decoder.GUID = D3BBF1E9-2427-450D-BBFF-B2843C1D44CC
simple_decoder.SRCS += ivfdec.h ivfdec.c

852
examples/vpx_temporal_svc_encoder.c
Просмотреть файл

@ -1,852 +0,0 @@
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
// This is an example demonstrating how to implement a multi-layer VPx
// encoding scheme based on temporal scalability for video applications
// that benefit from a scalable bitstream.
#include <assert.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "./vpx_config.h"
#include "../vpx_ports/vpx_timer.h"
#include "vpx/vp8cx.h"
#include "vpx/vpx_encoder.h"
#include "../tools_common.h"
#include "../video_writer.h"
static const char *exec_name;
void usage_exit(void) {
exit(EXIT_FAILURE);
}
// Denoiser states, for temporal denoising.
enum denoiserState {
kDenoiserOff,
kDenoiserOnYOnly,
kDenoiserOnYUV,
kDenoiserOnYUVAggressive,
kDenoiserOnAdaptive
};
static int mode_to_num_layers[13] = {1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3, 3};
// For rate control encoding stats.
struct RateControlMetrics {
// Number of input frames per layer.
int layer_input_frames[VPX_TS_MAX_LAYERS];
// Total (cumulative) number of encoded frames per layer.
int layer_tot_enc_frames[VPX_TS_MAX_LAYERS];
// Number of encoded non-key frames per layer.
int layer_enc_frames[VPX_TS_MAX_LAYERS];
// Framerate per layer layer (cumulative).
double layer_framerate[VPX_TS_MAX_LAYERS];
// Target average frame size per layer (per-frame-bandwidth per layer).
double layer_pfb[VPX_TS_MAX_LAYERS];
// Actual average frame size per layer.
double layer_avg_frame_size[VPX_TS_MAX_LAYERS];
// Average rate mismatch per layer (|target - actual| / target).
double layer_avg_rate_mismatch[VPX_TS_MAX_LAYERS];
// Actual encoding bitrate per layer (cumulative).
double layer_encoding_bitrate[VPX_TS_MAX_LAYERS];
// Average of the short-time encoder actual bitrate.
// TODO(marpan): Should we add these short-time stats for each layer?
double avg_st_encoding_bitrate;
// Variance of the short-time encoder actual bitrate.
double variance_st_encoding_bitrate;
// Window (number of frames) for computing short-timee encoding bitrate.
int window_size;
// Number of window measurements.
int window_count;
int layer_target_bitrate[VPX_MAX_LAYERS];
};
// Note: these rate control metrics assume only 1 key frame in the
// sequence (i.e., first frame only). So for temporal pattern# 7
// (which has key frame for every frame on base layer), the metrics
// computation will be off/wrong.
// TODO(marpan): Update these metrics to account for multiple key frames
// in the stream.
static void set_rate_control_metrics(struct RateControlMetrics *rc,
vpx_codec_enc_cfg_t *cfg) {
unsigned int i = 0;
// Set the layer (cumulative) framerate and the target layer (non-cumulative)
// per-frame-bandwidth, for the rate control encoding stats below.
const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
rc->layer_framerate[0] = framerate / cfg->ts_rate_decimator[0];
rc->layer_pfb[0] = 1000.0 * rc->layer_target_bitrate[0] /
rc->layer_framerate[0];
for (i = 0; i < cfg->ts_number_layers; ++i) {
if (i > 0) {
rc->layer_framerate[i] = framerate / cfg->ts_rate_decimator[i];
rc->layer_pfb[i] = 1000.0 *
(rc->layer_target_bitrate[i] - rc->layer_target_bitrate[i - 1]) /
(rc->layer_framerate[i] - rc->layer_framerate[i - 1]);
}
rc->layer_input_frames[i] = 0;
rc->layer_enc_frames[i] = 0;
rc->layer_tot_enc_frames[i] = 0;
rc->layer_encoding_bitrate[i] = 0.0;
rc->layer_avg_frame_size[i] = 0.0;
rc->layer_avg_rate_mismatch[i] = 0.0;
}
rc->window_count = 0;
rc->window_size = 15;
rc->avg_st_encoding_bitrate = 0.0;
rc->variance_st_encoding_bitrate = 0.0;
}
static void printout_rate_control_summary(struct RateControlMetrics *rc,
vpx_codec_enc_cfg_t *cfg,
int frame_cnt) {
unsigned int i = 0;
int tot_num_frames = 0;
double perc_fluctuation = 0.0;
printf("Total number of processed frames: %d\n\n", frame_cnt -1);
printf("Rate control layer stats for %d layer(s):\n\n",
cfg->ts_number_layers);
for (i = 0; i < cfg->ts_number_layers; ++i) {
const int num_dropped = (i > 0) ?
(rc->layer_input_frames[i] - rc->layer_enc_frames[i]) :
(rc->layer_input_frames[i] - rc->layer_enc_frames[i] - 1);
tot_num_frames += rc->layer_input_frames[i];
rc->layer_encoding_bitrate[i] = 0.001 * rc->layer_framerate[i] *
rc->layer_encoding_bitrate[i] / tot_num_frames;
rc->layer_avg_frame_size[i] = rc->layer_avg_frame_size[i] /
rc->layer_enc_frames[i];
rc->layer_avg_rate_mismatch[i] = 100.0 * rc->layer_avg_rate_mismatch[i] /
rc->layer_enc_frames[i];
printf("For layer#: %d \n", i);
printf("Bitrate (target vs actual): %d %f \n", rc->layer_target_bitrate[i],
rc->layer_encoding_bitrate[i]);
printf("Average frame size (target vs actual): %f %f \n", rc->layer_pfb[i],
rc->layer_avg_frame_size[i]);
printf("Average rate_mismatch: %f \n", rc->layer_avg_rate_mismatch[i]);
printf("Number of input frames, encoded (non-key) frames, "
"and perc dropped frames: %d %d %f \n", rc->layer_input_frames[i],
rc->layer_enc_frames[i],
100.0 * num_dropped / rc->layer_input_frames[i]);
printf("\n");
}
rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
rc->variance_st_encoding_bitrate =
rc->variance_st_encoding_bitrate / rc->window_count -
(rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
rc->avg_st_encoding_bitrate;
printf("Short-time stats, for window of %d frames: \n",rc->window_size);
printf("Average, rms-variance, and percent-fluct: %f %f %f \n",
rc->avg_st_encoding_bitrate,
sqrt(rc->variance_st_encoding_bitrate),
perc_fluctuation);
if ((frame_cnt - 1) != tot_num_frames)
die("Error: Number of input frames not equal to output! \n");
}
// Temporal scaling parameters:
// NOTE: The 3 prediction frames cannot be used interchangeably due to
// differences in the way they are handled throughout the code. The
// frames should be allocated to layers in the order LAST, GF, ARF.
// Other combinations work, but may produce slightly inferior results.
static void set_temporal_layer_pattern(int layering_mode,
vpx_codec_enc_cfg_t *cfg,
int *layer_flags,
int *flag_periodicity) {
switch (layering_mode) {
case 0: {
// 1-layer.
int ids[1] = {0};
cfg->ts_periodicity = 1;
*flag_periodicity = 1;
cfg->ts_number_layers = 1;
cfg->ts_rate_decimator[0] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// Update L only.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
break;
}
case 1: {
// 2-layers, 2-frame period.
int ids[2] = {0, 1};
cfg->ts_periodicity = 2;
*flag_periodicity = 2;
cfg->ts_number_layers = 2;
cfg->ts_rate_decimator[0] = 2;
cfg->ts_rate_decimator[1] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
#if 1
// 0=L, 1=GF, Intra-layer prediction enabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_REF_ARF;
#else
// 0=L, 1=GF, Intra-layer prediction disabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_LAST;
#endif
break;
}
case 2: {
// 2-layers, 3-frame period.
int ids[3] = {0, 1, 1};
cfg->ts_periodicity = 3;
*flag_periodicity = 3;
cfg->ts_number_layers = 2;
cfg->ts_rate_decimator[0] = 3;
cfg->ts_rate_decimator[1] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, Intra-layer prediction enabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[1] =
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
break;
}
case 3: {
// 3-layers, 6-frame period.
int ids[6] = {0, 2, 2, 1, 2, 2};
cfg->ts_periodicity = 6;
*flag_periodicity = 6;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 6;
cfg->ts_rate_decimator[1] = 3;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST;
layer_flags[1] =
layer_flags[2] =
layer_flags[4] =
layer_flags[5] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_LAST;
break;
}
case 4: {
// 3-layers, 4-frame period.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 4;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
layer_flags[1] =
layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
break;
}
case 5: {
// 3-layers, 4-frame period.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 4;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled in layer 1, disabled
// in layer 2.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
layer_flags[1] =
layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
break;
}
case 6: {
// 3-layers, 4-frame period.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 4;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
layer_flags[1] =
layer_flags[3] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
break;
}
case 7: {
// NOTE: Probably of academic interest only.
// 5-layers, 16-frame period.
int ids[16] = {0, 4, 3, 4, 2, 4, 3, 4, 1, 4, 3, 4, 2, 4, 3, 4};
cfg->ts_periodicity = 16;
*flag_periodicity = 16;
cfg->ts_number_layers = 5;
cfg->ts_rate_decimator[0] = 16;
cfg->ts_rate_decimator[1] = 8;
cfg->ts_rate_decimator[2] = 4;
cfg->ts_rate_decimator[3] = 2;
cfg->ts_rate_decimator[4] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
layer_flags[0] = VPX_EFLAG_FORCE_KF;
layer_flags[1] =
layer_flags[3] =
layer_flags[5] =
layer_flags[7] =
layer_flags[9] =
layer_flags[11] =
layer_flags[13] =
layer_flags[15] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
layer_flags[2] =
layer_flags[6] =
layer_flags[10] =
layer_flags[14] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF;
layer_flags[4] =
layer_flags[12] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_UPD_ARF;
layer_flags[8] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF;
break;
}
case 8: {
// 2-layers, with sync point at first frame of layer 1.
int ids[2] = {0, 1};
cfg->ts_periodicity = 2;
*flag_periodicity = 8;
cfg->ts_number_layers = 2;
cfg->ts_rate_decimator[0] = 2;
cfg->ts_rate_decimator[1] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF.
// ARF is used as predictor for all frames, and is only updated on
// key frame. Sync point every 8 frames.
// Layer 0: predict from L and ARF, update L and G.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_UPD_ARF;
// Layer 1: sync point: predict from L and ARF, and update G.
layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
// Layer 0, predict from L and ARF, update L.
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
// Layer 1: predict from L, G and ARF, and update G.
layer_flags[3] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ENTROPY;
// Layer 0.
layer_flags[4] = layer_flags[2];
// Layer 1.
layer_flags[5] = layer_flags[3];
// Layer 0.
layer_flags[6] = layer_flags[4];
// Layer 1.
layer_flags[7] = layer_flags[5];
break;
}
case 9: {
// 3-layers: Sync points for layer 1 and 2 every 8 frames.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 8;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ARF;
layer_flags[3] =
layer_flags[5] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
layer_flags[4] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[6] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
layer_flags[7] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_ENTROPY;
break;
}
case 10: {
// 3-layers structure where ARF is used as predictor for all frames,
// and is only updated on key frame.
// Sync points for layer 1 and 2 every 8 frames.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 8;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF.
// Layer 0: predict from L and ARF; update L and G.
layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_REF_GF;
// Layer 2: sync point: predict from L and ARF; update none.
layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ENTROPY;
// Layer 1: sync point: predict from L and ARF; update G.
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST;
// Layer 2: predict from L, G, ARF; update none.
layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
// Layer 0: predict from L and ARF; update L.
layer_flags[4] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_REF_GF;
// Layer 2: predict from L, G, ARF; update none.
layer_flags[5] = layer_flags[3];
// Layer 1: predict from L, G, ARF; update G.
layer_flags[6] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
// Layer 2: predict from L, G, ARF; update none.
layer_flags[7] = layer_flags[3];
break;
}
case 11: {
// 3-layers structure with one reference frame.
// This works same as temporal_layering_mode 3.
// This was added to compare with vp9_spatial_svc_encoder.
// 3-layers, 4-frame period.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 4;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
layer_flags[0] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
layer_flags[3] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
break;
}
case 12:
default: {
// 3-layers structure as in case 10, but no sync/refresh points for
// layer 1 and 2.
int ids[4] = {0, 2, 1, 2};
cfg->ts_periodicity = 4;
*flag_periodicity = 8;
cfg->ts_number_layers = 3;
cfg->ts_rate_decimator[0] = 4;
cfg->ts_rate_decimator[1] = 2;
cfg->ts_rate_decimator[2] = 1;
memcpy(cfg->ts_layer_id, ids, sizeof(ids));
// 0=L, 1=GF, 2=ARF.
// Layer 0: predict from L and ARF; update L.
layer_flags[0] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_REF_GF;
layer_flags[4] = layer_flags[0];
// Layer 1: predict from L, G, ARF; update G.
layer_flags[2] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
layer_flags[6] = layer_flags[2];
// Layer 2: predict from L, G, ARF; update none.
layer_flags[1] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
layer_flags[3] = layer_flags[1];
layer_flags[5] = layer_flags[1];
layer_flags[7] = layer_flags[1];
break;
}
}
}
int main(int argc, char **argv) {
VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL};
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_cnt = 0;
vpx_image_t raw;
vpx_codec_err_t res;
unsigned int width;
unsigned int height;
int speed;
int frame_avail;
int got_data;
int flags = 0;
unsigned int i;
int pts = 0; // PTS starts at 0.
int frame_duration = 1; // 1 timebase tick per frame.
int layering_mode = 0;
int layer_flags[VPX_TS_MAX_PERIODICITY] = {0};
int flag_periodicity = 1;
#if VPX_ENCODER_ABI_VERSION > (4 + VPX_CODEC_ABI_VERSION)
vpx_svc_layer_id_t layer_id = {0, 0};
#else
vpx_svc_layer_id_t layer_id = {0};
#endif
const VpxInterface *encoder = NULL;
FILE *infile = NULL;
struct RateControlMetrics rc;
int64_t cx_time = 0;
const int min_args_base = 11;
#if CONFIG_VP9_HIGHBITDEPTH
vpx_bit_depth_t bit_depth = VPX_BITS_8;
int input_bit_depth = 8;
const int min_args = min_args_base + 1;
#else
const int min_args = min_args_base;
#endif // CONFIG_VP9_HIGHBITDEPTH
double sum_bitrate = 0.0;
double sum_bitrate2 = 0.0;
double framerate = 30.0;
exec_name = argv[0];
// Check usage and arguments.
if (argc < min_args) {
#if CONFIG_VP9_HIGHBITDEPTH
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n", argv[0]);
#else
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n", argv[0]);
#endif // CONFIG_VP9_HIGHBITDEPTH
}
encoder = get_vpx_encoder_by_name(argv[3]);
if (!encoder)
die("Unsupported codec.");
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
width = strtol(argv[4], NULL, 0);
height = strtol(argv[5], NULL, 0);
if (width < 16 || width % 2 || height < 16 || height % 2) {
die("Invalid resolution: %d x %d", width, height);
}
layering_mode = strtol(argv[10], NULL, 0);
if (layering_mode < 0 || layering_mode > 13) {
die("Invalid layering mode (0..12) %s", argv[10]);
}
if (argc != min_args + mode_to_num_layers[layering_mode]) {
die("Invalid number of arguments");
}
#if CONFIG_VP9_HIGHBITDEPTH
switch (strtol(argv[argc-1], NULL, 0)) {
case 8:
bit_depth = VPX_BITS_8;
input_bit_depth = 8;
break;
case 10:
bit_depth = VPX_BITS_10;
input_bit_depth = 10;
break;
case 12:
bit_depth = VPX_BITS_12;
input_bit_depth = 12;
break;
default:
die("Invalid bit depth (8, 10, 12) %s", argv[argc-1]);
}
if (!vpx_img_alloc(&raw,
bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 :
VPX_IMG_FMT_I42016,
width, height, 32)) {
die("Failed to allocate image", width, height);
}
#else
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
die("Failed to allocate image", width, height);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Populate encoder configuration.
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) {
printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
return EXIT_FAILURE;
}
// Update the default configuration with our settings.
cfg.g_w = width;
cfg.g_h = height;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth != VPX_BITS_8) {
cfg.g_bit_depth = bit_depth;
cfg.g_input_bit_depth = input_bit_depth;
cfg.g_profile = 2;
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Timebase format e.g. 30fps: numerator=1, demoninator = 30.
cfg.g_timebase.num = strtol(argv[6], NULL, 0);
cfg.g_timebase.den = strtol(argv[7], NULL, 0);
speed = strtol(argv[8], NULL, 0);
if (speed < 0) {
die("Invalid speed setting: must be positive");
}
for (i = min_args_base;
(int)i < min_args_base + mode_to_num_layers[layering_mode];
++i) {
rc.layer_target_bitrate[i - 11] = strtol(argv[i], NULL, 0);
if (strncmp(encoder->name, "vp8", 3) == 0)
cfg.ts_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
else if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.layer_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
}
// Real time parameters.
cfg.rc_dropframe_thresh = strtol(argv[9], NULL, 0);
cfg.rc_end_usage = VPX_CBR;
cfg.rc_min_quantizer = 2;
cfg.rc_max_quantizer = 56;
if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.rc_max_quantizer = 52;
cfg.rc_undershoot_pct = 50;
cfg.rc_overshoot_pct = 50;
cfg.rc_buf_initial_sz = 500;
cfg.rc_buf_optimal_sz = 600;
cfg.rc_buf_sz = 1000;
// Disable dynamic resizing by default.
cfg.rc_resize_allowed = 0;
// Use 1 thread as default.
cfg.g_threads = 1;
// Enable error resilient mode.
cfg.g_error_resilient = 1;
cfg.g_lag_in_frames = 0;
cfg.kf_mode = VPX_KF_AUTO;
// Disable automatic keyframe placement.
cfg.kf_min_dist = cfg.kf_max_dist = 3000;
cfg.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
set_temporal_layer_pattern(layering_mode,
&cfg,
layer_flags,
&flag_periodicity);
set_rate_control_metrics(&rc, &cfg);
// Target bandwidth for the whole stream.
// Set to layer_target_bitrate for highest layer (total bitrate).
cfg.rc_target_bitrate = rc.layer_target_bitrate[cfg.ts_number_layers - 1];
// Open input file.
if (!(infile = fopen(argv[1], "rb"))) {
die("Failed to open %s for reading", argv[1]);
}
framerate = cfg.g_timebase.den / cfg.g_timebase.num;
// Open an output file for each stream.
for (i = 0; i < cfg.ts_number_layers; ++i) {
char file_name[PATH_MAX];
VpxVideoInfo info;
info.codec_fourcc = encoder->fourcc;
info.frame_width = cfg.g_w;
info.frame_height = cfg.g_h;
info.time_base.numerator = cfg.g_timebase.num;
info.time_base.denominator = cfg.g_timebase.den;
snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
if (!outfile[i])
die("Failed to open %s for writing", file_name);
assert(outfile[i] != NULL);
}
// No spatial layers in this encoder.
cfg.ss_number_layers = 1;
// Initialize codec.
#if CONFIG_VP9_HIGHBITDEPTH
if (vpx_codec_enc_init(
&codec, encoder->codec_interface(), &cfg,
bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
#else
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
#endif // CONFIG_VP9_HIGHBITDEPTH
die_codec(&codec, "Failed to initialize encoder");
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kDenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
} else if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_svc_extra_cfg_t svc_params;
vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kDenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP9E_SET_TUNE_CONTENT, 0);
vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (cfg.g_threads >> 1));
if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1: 0))
die_codec(&codec, "Failed to set SVC");
for (i = 0; i < cfg.ts_number_layers; ++i) {
svc_params.max_quantizers[i] = cfg.rc_max_quantizer;
svc_params.min_quantizers[i] = cfg.rc_min_quantizer;
}
svc_params.scaling_factor_num[0] = cfg.g_h;
svc_params.scaling_factor_den[0] = cfg.g_h;
vpx_codec_control(&codec, VP9E_SET_SVC_PARAMETERS, &svc_params);
}
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_SCREEN_CONTENT_MODE, 0);
}
vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1);
// This controls the maximum target size of the key frame.
// For generating smaller key frames, use a smaller max_intra_size_pct
// value, like 100 or 200.
{
const int max_intra_size_pct = 900;
vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT,
max_intra_size_pct);
}
frame_avail = 1;
while (frame_avail || got_data) {
struct vpx_usec_timer timer;
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *pkt;
#if VPX_ENCODER_ABI_VERSION > (4 + VPX_CODEC_ABI_VERSION)
// Update the temporal layer_id. No spatial layers in this test.
layer_id.spatial_layer_id = 0;
#endif
layer_id.temporal_layer_id =
cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
} else if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_TEMPORAL_LAYER_ID,
layer_id.temporal_layer_id);
}
flags = layer_flags[frame_cnt % flag_periodicity];
if (layering_mode == 0)
flags = 0;
frame_avail = vpx_img_read(&raw, infile);
if (frame_avail)
++rc.layer_input_frames[layer_id.temporal_layer_id];
vpx_usec_timer_start(&timer);
if (vpx_codec_encode(&codec, frame_avail? &raw : NULL, pts, 1, flags,
VPX_DL_REALTIME)) {
die_codec(&codec, "Failed to encode frame");
}
vpx_usec_timer_mark(&timer);
cx_time += vpx_usec_timer_elapsed(&timer);
// Reset KF flag.
if (layering_mode != 7) {
layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
}
got_data = 0;
while ( (pkt = vpx_codec_get_cx_data(&codec, &iter)) ) {
got_data = 1;
switch (pkt->kind) {
case VPX_CODEC_CX_FRAME_PKT:
for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
i < cfg.ts_number_layers; ++i) {
vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
pkt->data.frame.sz, pts);
++rc.layer_tot_enc_frames[i];
rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
// Keep count of rate control stats per layer (for non-key frames).
if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
!(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
rc.layer_avg_rate_mismatch[i] +=
fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
rc.layer_pfb[i];
++rc.layer_enc_frames[i];
}
}
// Update for short-time encoding bitrate states, for moving window
// of size rc->window, shifted by rc->window / 2.
// Ignore first window segment, due to key frame.
if (frame_cnt > rc.window_size) {
sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate / rc.window_size) *
(sum_bitrate / rc.window_size);
sum_bitrate = 0.0;
}
}
// Second shifted window.
if (frame_cnt > rc.window_size + rc.window_size / 2) {
sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt > 2 * rc.window_size &&
frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate2 / rc.window_size) *
(sum_bitrate2 / rc.window_size);
sum_bitrate2 = 0.0;
}
}
break;
default:
break;
}
}
++frame_cnt;
pts += frame_duration;
}
fclose(infile);
printout_rate_control_summary(&rc, &cfg, frame_cnt);
printf("\n");
printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
frame_cnt,
1000 * (float)cx_time / (double)(frame_cnt * 1000000),
1000000 * (double)frame_cnt / (double)cx_time);
if (vpx_codec_destroy(&codec))
die_codec(&codec, "Failed to destroy codec");
// Try to rewrite the output file headers with the actual frame count.
for (i = 0; i < cfg.ts_number_layers; ++i)
vpx_video_writer_close(outfile[i]);
vpx_img_free(&raw);
return EXIT_SUCCESS;
}

6
libs.mk
Просмотреть файл

@ -66,7 +66,6 @@ ifeq ($(CONFIG_VP10_ENCODER),yes)
CODEC_EXPORTS-yes += $(addprefix $(VP10_PREFIX),$(VP10_CX_EXPORTS))
CODEC_SRCS-yes += $(VP10_PREFIX)vp10cx.mk vpx/vp8.h vpx/vp8cx.h
INSTALL-LIBS-yes += include/vpx/vp8.h include/vpx/vp8cx.h
INSTALL-LIBS-$(CONFIG_SPATIAL_SVC) += include/vpx/svc_context.h
INSTALL_MAPS += include/vpx/% $(SRC_PATH_BARE)/$(VP10_PREFIX)/%
CODEC_DOC_SRCS += vpx/vp8.h vpx/vp8cx.h
CODEC_DOC_SECTIONS += vp9 vp9_encoder
@ -127,9 +126,6 @@ INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += third_party/x86inc/x86inc.asm
endif
CODEC_EXPORTS-yes += vpx/exports_com
CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_enc
ifeq ($(CONFIG_SPATIAL_SVC),yes)
CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_spatial_svc
endif
CODEC_EXPORTS-$(CONFIG_DECODERS) += vpx/exports_dec
INSTALL-LIBS-yes += include/vpx/vpx_codec.h
@ -265,7 +261,7 @@ libvpx.def: $(call enabled,CODEC_EXPORTS)
$(qexec)echo LIBRARY $(LIBVPX_SO:.dll=) INITINSTANCE TERMINSTANCE > $@
$(qexec)echo "DATA MULTIPLE NONSHARED" >> $@
$(qexec)echo "EXPORTS" >> $@
$(qexec)awk '!/vpx_svc_*/ {print "_"$$2}' $^ >>$@
$(qexec)awk '{print "_"$$2}' $^ >>$@
CLEAN-OBJS += libvpx.def
libvpx_dll.a: $(LIBVPX_SO)

838
test/datarate_test.cc
Просмотреть файл

@ -19,198 +19,13 @@
namespace {
class DatarateTestLarge : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, int> {
public:
DatarateTestLarge() : EncoderTest(GET_PARAM(0)) {}
protected:
virtual ~DatarateTestLarge() {}
protected:
virtual void SetUp() {
InitializeConfig();
SetMode(GET_PARAM(1));
ResetModel();
}
virtual void ResetModel() {
last_pts_ = 0;
bits_in_buffer_model_ = cfg_.rc_target_bitrate * cfg_.rc_buf_initial_sz;
frame_number_ = 0;
first_drop_ = 0;
bits_total_ = 0;
duration_ = 0.0;
denoiser_offon_test_ = 0;
denoiser_offon_period_ = -1;
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
::libvpx_test::Encoder *encoder) {
if (video->frame() == 0)
encoder->Control(VP8E_SET_NOISE_SENSITIVITY, denoiser_on_);
if (denoiser_offon_test_) {
ASSERT_GT(denoiser_offon_period_, 0)
<< "denoiser_offon_period_ is not positive.";
if ((video->frame() + 1) % denoiser_offon_period_ == 0) {
// Flip denoiser_on_ periodically
denoiser_on_ ^= 1;
}
encoder->Control(VP8E_SET_NOISE_SENSITIVITY, denoiser_on_);
}
const vpx_rational_t tb = video->timebase();
timebase_ = static_cast<double>(tb.num) / tb.den;
duration_ = 0;
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
// Time since last timestamp = duration.
vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
// TODO(jimbankoski): Remove these lines when the issue:
// http://code.google.com/p/webm/issues/detail?id=496 is fixed.
// For now the codec assumes buffer starts at starting buffer rate
// plus one frame's time.
if (last_pts_ == 0)
duration = 1;
// Add to the buffer the bits we'd expect from a constant bitrate server.
bits_in_buffer_model_ += static_cast<int64_t>(
duration * timebase_ * cfg_.rc_target_bitrate * 1000);
/* Test the buffer model here before subtracting the frame. Do so because
* the way the leaky bucket model works in libvpx is to allow the buffer to
* empty - and then stop showing frames until we've got enough bits to
* show one. As noted in comment below (issue 495), this does not currently
* apply to key frames. For now exclude key frames in condition below. */
const bool key_frame = (pkt->data.frame.flags & VPX_FRAME_IS_KEY)
? true: false;
if (!key_frame) {
ASSERT_GE(bits_in_buffer_model_, 0) << "Buffer Underrun at frame "
<< pkt->data.frame.pts;
}
const int64_t frame_size_in_bits = pkt->data.frame.sz * 8;
// Subtract from the buffer the bits associated with a played back frame.
bits_in_buffer_model_ -= frame_size_in_bits;
// Update the running total of bits for end of test datarate checks.
bits_total_ += frame_size_in_bits;
// If first drop not set and we have a drop set it to this time.
if (!first_drop_ && duration > 1)
first_drop_ = last_pts_ + 1;
// Update the most recent pts.
last_pts_ = pkt->data.frame.pts;
// We update this so that we can calculate the datarate minus the last
// frame encoded in the file.
bits_in_last_frame_ = frame_size_in_bits;
++frame_number_;
}
virtual void EndPassHook(void) {
if (bits_total_) {
const double file_size_in_kb = bits_total_ / 1000.; // bits per kilobit
duration_ = (last_pts_ + 1) * timebase_;
// Effective file datarate includes the time spent prebuffering.
effective_datarate_ = (bits_total_ - bits_in_last_frame_) / 1000.0
/ (cfg_.rc_buf_initial_sz / 1000.0 + duration_);
file_datarate_ = file_size_in_kb / duration_;
}
}
vpx_codec_pts_t last_pts_;
int64_t bits_in_buffer_model_;
double timebase_;
int frame_number_;
vpx_codec_pts_t first_drop_;
int64_t bits_total_;
double duration_;
double file_datarate_;
double effective_datarate_;
int64_t bits_in_last_frame_;
int denoiser_on_;
int denoiser_offon_test_;
int denoiser_offon_period_;
};
TEST_P(DatarateTestLarge, BasicBufferModel) {
denoiser_on_ = 0;
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_dropframe_thresh = 1;
cfg_.rc_max_quantizer = 56;
cfg_.rc_end_usage = VPX_CBR;
// 2 pass cbr datarate control has a bug hidden by the small # of
// frames selected in this encode. The problem is that even if the buffer is
// negative we produce a keyframe on a cutscene. Ignoring datarate
// constraints
// TODO(jimbankoski): ( Fix when issue
// http://code.google.com/p/webm/issues/detail?id=495 is addressed. )
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 140);
// There is an issue for low bitrates in real-time mode, where the
// effective_datarate slightly overshoots the target bitrate.
// This is same the issue as noted about (#495).
// TODO(jimbankoski/marpan): Update test to run for lower bitrates (< 100),
// when the issue is resolved.
for (int i = 100; i < 800; i += 200) {
cfg_.rc_target_bitrate = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, effective_datarate_ * 0.95)
<< " The datarate for the file exceeds the target!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.3)
<< " The datarate for the file missed the target!";
}
}
TEST_P(DatarateTestLarge, ChangingDropFrameThresh) {
denoiser_on_ = 0;
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_max_quantizer = 36;
cfg_.rc_end_usage = VPX_CBR;
cfg_.rc_target_bitrate = 200;
cfg_.kf_mode = VPX_KF_DISABLED;
const int frame_count = 40;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, frame_count);
// Here we check that the first dropped frame gets earlier and earlier
// as the drop frame threshold is increased.
const int kDropFrameThreshTestStep = 30;
vpx_codec_pts_t last_drop = frame_count;
for (int i = 1; i < 91; i += kDropFrameThreshTestStep) {
cfg_.rc_dropframe_thresh = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_LE(first_drop_, last_drop)
<< " The first dropped frame for drop_thresh " << i
<< " > first dropped frame for drop_thresh "
<< i - kDropFrameThreshTestStep;
last_drop = first_drop_;
}
}
class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, int> {
public:
DatarateTestVP9Large() : EncoderTest(GET_PARAM(0)) {}
protected:
virtual ~DatarateTestVP9Large() {}
virtual void SetUp() {
InitializeConfig();
SetMode(GET_PARAM(1));
@ -227,77 +42,11 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
num_drops_ = 0;
// Denoiser is off by default.
denoiser_on_ = 0;
// For testing up to 3 layers.
for (int i = 0; i < 3; ++i) {
bits_total_[i] = 0;
}
bits_total_ = 0;
denoiser_offon_test_ = 0;
denoiser_offon_period_ = -1;
}
//
// Frame flags and layer id for temporal layers.
//
// For two layers, test pattern is:
// 1 3
// 0 2 .....
// For three layers, test pattern is:
// 1 3 5 7
// 2 6
// 0 4 ....
// LAST is always update on base/layer 0, GOLDEN is updated on layer 1.
// For this 3 layer example, the 2nd enhancement layer (layer 2) does not
// update any reference frames.
int SetFrameFlags(int frame_num, int num_temp_layers) {
int frame_flags = 0;
if (num_temp_layers == 2) {
if (frame_num % 2 == 0) {
// Layer 0: predict from L and ARF, update L.
frame_flags = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
} else {
// Layer 1: predict from L, G and ARF, and update G.
frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ENTROPY;
}
} else if (num_temp_layers == 3) {
if (frame_num % 4 == 0) {
// Layer 0: predict from L and ARF; update L.
frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_REF_GF;
} else if ((frame_num - 2) % 4 == 0) {
// Layer 1: predict from L, G, ARF; update G.
frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
} else if ((frame_num - 1) % 2 == 0) {
// Layer 2: predict from L, G, ARF; update none.
frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST;
}
}
return frame_flags;
}
int SetLayerId(int frame_num, int num_temp_layers) {
int layer_id = 0;
if (num_temp_layers == 2) {
if (frame_num % 2 == 0) {
layer_id = 0;
} else {
layer_id = 1;
}
} else if (num_temp_layers == 3) {
if (frame_num % 4 == 0) {
layer_id = 0;
} else if ((frame_num - 2) % 4 == 0) {
layer_id = 1;
} else if ((frame_num - 1) % 2 == 0) {
layer_id = 2;
}
}
return layer_id;
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
::libvpx_test::Encoder *encoder) {
if (video->frame() == 0)
@ -314,23 +63,11 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
encoder->Control(VP9E_SET_NOISE_SENSITIVITY, denoiser_on_);
if (cfg_.ts_number_layers > 1) {
if (video->frame() == 0) {
encoder->Control(VP9E_SET_SVC, 1);
}
vpx_svc_layer_id_t layer_id;
layer_id.spatial_layer_id = 0;
frame_flags_ = SetFrameFlags(video->frame(), cfg_.ts_number_layers);
layer_id.temporal_layer_id = SetLayerId(video->frame(),
cfg_.ts_number_layers);
encoder->Control(VP9E_SET_SVC_LAYER_ID, &layer_id);
}
const vpx_rational_t tb = video->timebase();
timebase_ = static_cast<double>(tb.num) / tb.den;
duration_ = 0;
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
// Time since last timestamp = duration.
vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
@ -346,8 +83,6 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
tot_frame_number_ += static_cast<int>(duration - 1);
}
int layer = SetLayerId(tot_frame_number_, cfg_.ts_number_layers);
// Add to the buffer the bits we'd expect from a constant bitrate server.
bits_in_buffer_model_ += static_cast<int64_t>(
duration * timebase_ * cfg_.rc_target_bitrate * 1000);
@ -358,11 +93,8 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
const size_t frame_size_in_bits = pkt->data.frame.sz * 8;
// Update the total encoded bits. For temporal layers, update the cumulative
// encoded bits per layer.
for (int i = layer; i < static_cast<int>(cfg_.ts_number_layers); ++i) {
bits_total_[i] += frame_size_in_bits;
}
// Update the total encoded bits.
bits_total_ += frame_size_in_bits;
// Update the most recent pts.
last_pts_ = pkt->data.frame.pts;
@ -371,23 +103,18 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
}
virtual void EndPassHook(void) {
for (int layer = 0; layer < static_cast<int>(cfg_.ts_number_layers);
++layer) {
duration_ = (last_pts_ + 1) * timebase_;
if (bits_total_[layer]) {
// Effective file datarate:
effective_datarate_[layer] = (bits_total_[layer] / 1000.0) / duration_;
}
}
duration_ = (last_pts_ + 1) * timebase_;
// Effective file datarate:
effective_datarate_ = (bits_total_ / 1000.0) / duration_;
}
vpx_codec_pts_t last_pts_;
double timebase_;
int frame_number_; // Counter for number of non-dropped/encoded frames.
int tot_frame_number_; // Counter for total number of input frames.
int64_t bits_total_[3];
int64_t bits_total_;
double duration_;
double effective_datarate_[3];
double effective_datarate_;
int set_cpu_used_;
int64_t bits_in_buffer_model_;
vpx_codec_pts_t first_drop_;
@ -398,7 +125,7 @@ class DatarateTestVP9Large : public ::libvpx_test::EncoderTest,
};
// Check basic rate targeting for VBR mode.
TEST_P(DatarateTestVP9Large, BasicRateTargetingVBR) {
TEST_P(DatarateTestLarge, BasicRateTargetingVBR) {
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.g_error_resilient = 0;
@ -411,15 +138,15 @@ TEST_P(DatarateTestVP9Large, BasicRateTargetingVBR) {
cfg_.rc_target_bitrate = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(effective_datarate_[0], cfg_.rc_target_bitrate * 0.75)
ASSERT_GE(effective_datarate_, cfg_.rc_target_bitrate * 0.75)
<< " The datarate for the file is lower than target by too much!";
ASSERT_LE(effective_datarate_[0], cfg_.rc_target_bitrate * 1.25)
ASSERT_LE(effective_datarate_, cfg_.rc_target_bitrate * 1.25)
<< " The datarate for the file is greater than target by too much!";
}
}
// Check basic rate targeting for CBR,
TEST_P(DatarateTestVP9Large, BasicRateTargeting) {
TEST_P(DatarateTestLarge, BasicRateTargeting) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
@ -435,15 +162,15 @@ TEST_P(DatarateTestVP9Large, BasicRateTargeting) {
cfg_.rc_target_bitrate = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(effective_datarate_[0], cfg_.rc_target_bitrate * 0.85)
ASSERT_GE(effective_datarate_, cfg_.rc_target_bitrate * 0.85)
<< " The datarate for the file is lower than target by too much!";
ASSERT_LE(effective_datarate_[0], cfg_.rc_target_bitrate * 1.15)
ASSERT_LE(effective_datarate_, cfg_.rc_target_bitrate * 1.15)
<< " The datarate for the file is greater than target by too much!";
}
}
// Check basic rate targeting for CBR.
TEST_P(DatarateTestVP9Large, BasicRateTargeting444) {
TEST_P(DatarateTestLarge, BasicRateTargeting444) {
::libvpx_test::Y4mVideoSource video("rush_hour_444.y4m", 0, 140);
cfg_.g_profile = 1;
@ -462,10 +189,10 @@ TEST_P(DatarateTestVP9Large, BasicRateTargeting444) {
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(static_cast<double>(cfg_.rc_target_bitrate),
effective_datarate_[0] * 0.85)
effective_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(static_cast<double>(cfg_.rc_target_bitrate),
effective_datarate_[0] * 1.15)
effective_datarate_ * 1.15)
<< " The datarate for the file missed the target!"
<< cfg_.rc_target_bitrate << " "<< effective_datarate_;
}
@ -475,7 +202,7 @@ TEST_P(DatarateTestVP9Large, BasicRateTargeting444) {
// as the drop frame threshold is increased, and (2) that the total number of
// frame drops does not decrease as we increase frame drop threshold.
// Use a lower qp-max to force some frame drops.
TEST_P(DatarateTestVP9Large, ChangingDropFrameThresh) {
TEST_P(DatarateTestLarge, ChangingDropFrameThresh) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
@ -501,9 +228,9 @@ TEST_P(DatarateTestVP9Large, ChangingDropFrameThresh) {
cfg_.rc_dropframe_thresh = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(effective_datarate_[0], cfg_.rc_target_bitrate * 0.85)
ASSERT_GE(effective_datarate_, cfg_.rc_target_bitrate * 0.85)
<< " The datarate for the file is lower than target by too much!";
ASSERT_LE(effective_datarate_[0], cfg_.rc_target_bitrate * 1.15)
ASSERT_LE(effective_datarate_, cfg_.rc_target_bitrate * 1.15)
<< " The datarate for the file is greater than target by too much!";
ASSERT_LE(first_drop_, last_drop)
<< " The first dropped frame for drop_thresh " << i
@ -518,521 +245,8 @@ TEST_P(DatarateTestVP9Large, ChangingDropFrameThresh) {
}
}
// Check basic rate targeting for 2 temporal layers.
TEST_P(DatarateTestVP9Large, BasicRateTargeting2TemporalLayers) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_dropframe_thresh = 1;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
// 2 Temporal layers, no spatial layers: Framerate decimation (2, 1).
cfg_.ss_number_layers = 1;
cfg_.ts_number_layers = 2;
cfg_.ts_rate_decimator[0] = 2;
cfg_.ts_rate_decimator[1] = 1;
cfg_.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
if (deadline_ == VPX_DL_REALTIME)
cfg_.g_error_resilient = 1;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
for (int i = 200; i <= 800; i += 200) {
cfg_.rc_target_bitrate = i;
ResetModel();
// 60-40 bitrate allocation for 2 temporal layers.
cfg_.layer_target_bitrate[0] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.layer_target_bitrate[1] = cfg_.rc_target_bitrate;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {
ASSERT_GE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 0.85)
<< " The datarate for the file is lower than target by too much, "
"for layer: " << j;
ASSERT_LE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 1.15)
<< " The datarate for the file is greater than target by too much, "
"for layer: " << j;
}
}
}
// Check basic rate targeting for 3 temporal layers.
TEST_P(DatarateTestVP9Large, BasicRateTargeting3TemporalLayers) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_dropframe_thresh = 1;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
// 3 Temporal layers, no spatial layers: Framerate decimation (4, 2, 1).
cfg_.ss_number_layers = 1;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
for (int i = 200; i <= 800; i += 200) {
cfg_.rc_target_bitrate = i;
ResetModel();
// 40-20-40 bitrate allocation for 3 temporal layers.
cfg_.layer_target_bitrate[0] = 40 * cfg_.rc_target_bitrate / 100;
cfg_.layer_target_bitrate[1] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.layer_target_bitrate[2] = cfg_.rc_target_bitrate;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {
// TODO(yaowu): Work out more stable rc control strategy and
// Adjust the thresholds to be tighter than .75.
ASSERT_GE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 0.75)
<< " The datarate for the file is lower than target by too much, "
"for layer: " << j;
// TODO(yaowu): Work out more stable rc control strategy and
// Adjust the thresholds to be tighter than 1.25.
ASSERT_LE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 1.25)
<< " The datarate for the file is greater than target by too much, "
"for layer: " << j;
}
}
}
// Check basic rate targeting for 3 temporal layers, with frame dropping.
// Only for one (low) bitrate with lower max_quantizer, and somewhat higher
// frame drop threshold, to force frame dropping.
TEST_P(DatarateTestVP9Large, BasicRateTargeting3TemporalLayersFrameDropping) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
// Set frame drop threshold and rc_max_quantizer to force some frame drops.
cfg_.rc_dropframe_thresh = 20;
cfg_.rc_max_quantizer = 45;
cfg_.rc_min_quantizer = 0;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
// 3 Temporal layers, no spatial layers: Framerate decimation (4, 2, 1).
cfg_.ss_number_layers = 1;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
cfg_.rc_target_bitrate = 200;
ResetModel();
// 40-20-40 bitrate allocation for 3 temporal layers.
cfg_.layer_target_bitrate[0] = 40 * cfg_.rc_target_bitrate / 100;
cfg_.layer_target_bitrate[1] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.layer_target_bitrate[2] = cfg_.rc_target_bitrate;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {
ASSERT_GE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 0.85)
<< " The datarate for the file is lower than target by too much, "
"for layer: " << j;
ASSERT_LE(effective_datarate_[j], cfg_.layer_target_bitrate[j] * 1.15)
<< " The datarate for the file is greater than target by too much, "
"for layer: " << j;
// Expect some frame drops in this test: for this 200 frames test,
// expect at least 10% and not more than 60% drops.
ASSERT_GE(num_drops_, 20);
ASSERT_LE(num_drops_, 130);
}
}
class DatarateOnePassCbrSvc : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, int> {
public:
DatarateOnePassCbrSvc() : EncoderTest(GET_PARAM(0)) {}
virtual ~DatarateOnePassCbrSvc() {}
protected:
virtual void SetUp() {
InitializeConfig();
SetMode(GET_PARAM(1));
speed_setting_ = GET_PARAM(2);
ResetModel();
}
virtual void ResetModel() {
last_pts_ = 0;
bits_in_buffer_model_ = cfg_.rc_target_bitrate * cfg_.rc_buf_initial_sz;
frame_number_ = 0;
first_drop_ = 0;
bits_total_ = 0;
duration_ = 0.0;
mismatch_psnr_ = 0.0;
mismatch_nframes_ = 0;
}
virtual void BeginPassHook(unsigned int /*pass*/) {
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
::libvpx_test::Encoder *encoder) {
if (video->frame() == 0) {
int i;
for (i = 0; i < VPX_MAX_LAYERS; ++i) {
svc_params_.max_quantizers[i] = 63;
svc_params_.min_quantizers[i] = 0;
}
encoder->Control(VP9E_SET_SVC, 1);
encoder->Control(VP9E_SET_SVC_PARAMETERS, &svc_params_);
encoder->Control(VP8E_SET_CPUUSED, speed_setting_);
encoder->Control(VP9E_SET_TILE_COLUMNS, 0);
encoder->Control(VP8E_SET_MAX_INTRA_BITRATE_PCT, 300);
encoder->Control(VP9E_SET_TILE_COLUMNS, (cfg_.g_threads >> 1));
}
const vpx_rational_t tb = video->timebase();
timebase_ = static_cast<double>(tb.num) / tb.den;
duration_ = 0;
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
if (last_pts_ == 0)
duration = 1;
bits_in_buffer_model_ += static_cast<int64_t>(
duration * timebase_ * cfg_.rc_target_bitrate * 1000);
const bool key_frame = (pkt->data.frame.flags & VPX_FRAME_IS_KEY)
? true: false;
if (!key_frame) {
ASSERT_GE(bits_in_buffer_model_, 0) << "Buffer Underrun at frame "
<< pkt->data.frame.pts;
}
const size_t frame_size_in_bits = pkt->data.frame.sz * 8;
bits_in_buffer_model_ -= frame_size_in_bits;
bits_total_ += frame_size_in_bits;
if (!first_drop_ && duration > 1)
first_drop_ = last_pts_ + 1;
last_pts_ = pkt->data.frame.pts;
bits_in_last_frame_ = frame_size_in_bits;
++frame_number_;
}
virtual void EndPassHook(void) {
if (bits_total_) {
const double file_size_in_kb = bits_total_ / 1000.; // bits per kilobit
duration_ = (last_pts_ + 1) * timebase_;
file_datarate_ = file_size_in_kb / duration_;
}
}
virtual void MismatchHook(const vpx_image_t *img1,
const vpx_image_t *img2) {
double mismatch_psnr = compute_psnr(img1, img2);
mismatch_psnr_ += mismatch_psnr;
++mismatch_nframes_;
}
unsigned int GetMismatchFrames() {
return mismatch_nframes_;
}
vpx_codec_pts_t last_pts_;
int64_t bits_in_buffer_model_;
double timebase_;
int frame_number_;
vpx_codec_pts_t first_drop_;
int64_t bits_total_;
double duration_;
double file_datarate_;
size_t bits_in_last_frame_;
vpx_svc_extra_cfg_t svc_params_;
int speed_setting_;
double mismatch_psnr_;
int mismatch_nframes_;
};
static void assign_layer_bitrates(vpx_codec_enc_cfg_t *const enc_cfg,
const vpx_svc_extra_cfg_t *svc_params,
int spatial_layers,
int temporal_layers,
int temporal_layering_mode) {
int sl, spatial_layer_target;
float total = 0;
float alloc_ratio[VPX_MAX_LAYERS] = {0};
for (sl = 0; sl < spatial_layers; ++sl) {
if (svc_params->scaling_factor_den[sl] > 0) {
alloc_ratio[sl] = (float)(svc_params->scaling_factor_num[sl] *
1.0 / svc_params->scaling_factor_den[sl]);
total += alloc_ratio[sl];
}
}
for (sl = 0; sl < spatial_layers; ++sl) {
enc_cfg->ss_target_bitrate[sl] = spatial_layer_target =
(unsigned int)(enc_cfg->rc_target_bitrate *
alloc_ratio[sl] / total);
const int index = sl * temporal_layers;
if (temporal_layering_mode == 3) {
enc_cfg->layer_target_bitrate[index] =
spatial_layer_target >> 1;
enc_cfg->layer_target_bitrate[index + 1] =
(spatial_layer_target >> 1) + (spatial_layer_target >> 2);
enc_cfg->layer_target_bitrate[index + 2] =
spatial_layer_target;
} else if (temporal_layering_mode == 2) {
enc_cfg->layer_target_bitrate[index] =
spatial_layer_target * 2 / 3;
enc_cfg->layer_target_bitrate[index + 1] =
spatial_layer_target;
}
}
}
// Check basic rate targeting for 1 pass CBR SVC: 2 spatial layers and
// 3 temporal layers. Run CIF clip with 1 thread.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc2SpatialLayers) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 2;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 1;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 144;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 288;
svc_params_.scaling_factor_den[1] = 288;
cfg_.rc_dropframe_thresh = 10;
cfg_.kf_max_dist = 9999;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
// TODO(wonkap/marpan): Check that effective_datarate for each layer hits the
// layer target_bitrate.
for (int i = 200; i <= 800; i += 200) {
cfg_.rc_target_bitrate = i;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.15)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
}
// Check basic rate targeting for 1 pass CBR SVC: 2 spatial layers and 3
// temporal layers. Run CIF clip with 1 thread, and few short key frame periods.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc2SpatialLayersSmallKf) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 2;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 1;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 144;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 288;
svc_params_.scaling_factor_den[1] = 288;
cfg_.rc_dropframe_thresh = 10;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
cfg_.rc_target_bitrate = 400;
// For this 3 temporal layer case, pattern repeats every 4 frames, so choose
// 4 key neighboring key frame periods (so key frame will land on 0-2-1-2).
for (int j = 64; j <= 67; j++) {
cfg_.kf_max_dist = j;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.15)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
}
// Check basic rate targeting for 1 pass CBR SVC: 2 spatial layers and
// 3 temporal layers. Run HD clip with 4 threads.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc2SpatialLayers4threads) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 2;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 4;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 144;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 288;
svc_params_.scaling_factor_den[1] = 288;
cfg_.rc_dropframe_thresh = 10;
cfg_.kf_max_dist = 9999;
::libvpx_test::I420VideoSource video("niklas_1280_720_30.y4m", 1280, 720,
30, 1, 0, 300);
cfg_.rc_target_bitrate = 800;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.15)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
// Check basic rate targeting for 1 pass CBR SVC: 3 spatial layers and
// 3 temporal layers. Run CIF clip with 1 thread.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc3SpatialLayers) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 3;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 1;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 72;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 144;
svc_params_.scaling_factor_den[1] = 288;
svc_params_.scaling_factor_num[2] = 288;
svc_params_.scaling_factor_den[2] = 288;
cfg_.rc_dropframe_thresh = 10;
cfg_.kf_max_dist = 9999;
::libvpx_test::I420VideoSource video("niklas_1280_720_30.y4m", 1280, 720,
30, 1, 0, 300);
cfg_.rc_target_bitrate = 800;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.22)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
// Check basic rate targeting for 1 pass CBR SVC: 3 spatial layers and 3
// temporal layers. Run CIF clip with 1 thread, and few short key frame periods.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc3SpatialLayersSmallKf) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 3;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 1;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 72;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 144;
svc_params_.scaling_factor_den[1] = 288;
svc_params_.scaling_factor_num[2] = 288;
svc_params_.scaling_factor_den[2] = 288;
cfg_.rc_dropframe_thresh = 10;
::libvpx_test::I420VideoSource video("niklas_1280_720_30.y4m", 1280, 720,
30, 1, 0, 300);
cfg_.rc_target_bitrate = 800;
// For this 3 temporal layer case, pattern repeats every 4 frames, so choose
// 4 key neighboring key frame periods (so key frame will land on 0-2-1-2).
for (int j = 32; j <= 35; j++) {
cfg_.kf_max_dist = j;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.30)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
}
// Check basic rate targeting for 1 pass CBR SVC: 3 spatial layers and
// 3 temporal layers. Run HD clip with 4 threads.
TEST_P(DatarateOnePassCbrSvc, OnePassCbrSvc3SpatialLayers4threads) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
cfg_.g_lag_in_frames = 0;
cfg_.ss_number_layers = 3;
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.g_error_resilient = 1;
cfg_.g_threads = 4;
cfg_.temporal_layering_mode = 3;
svc_params_.scaling_factor_num[0] = 72;
svc_params_.scaling_factor_den[0] = 288;
svc_params_.scaling_factor_num[1] = 144;
svc_params_.scaling_factor_den[1] = 288;
svc_params_.scaling_factor_num[2] = 288;
svc_params_.scaling_factor_den[2] = 288;
cfg_.rc_dropframe_thresh = 10;
cfg_.kf_max_dist = 9999;
::libvpx_test::I420VideoSource video("niklas_1280_720_30.y4m", 1280, 720,
30, 1, 0, 300);
cfg_.rc_target_bitrate = 800;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, file_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.22)
<< " The datarate for the file is lower than the target by too much!";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
/* VP10 does not support multiple layers yet.
VP10_INSTANTIATE_TEST_CASE(DatarateOnePassCbrSvc,
::testing::Values(::libvpx_test::kRealTime),
::testing::Range(5, 8));
*/
VP10_INSTANTIATE_TEST_CASE(DatarateTestLarge,
::testing::Values(::libvpx_test::kOnePassGood,
::libvpx_test::kRealTime),
::testing::Range(2, 9));
} // namespace

9
test/encode_test_driver.h
Просмотреть файл

@ -134,15 +134,6 @@ class Encoder {
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
void Control(int ctrl_id, struct vpx_svc_layer_id *arg) {
const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
void Control(int ctrl_id, struct vpx_svc_parameters *arg) {
const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
#if CONFIG_VP10_ENCODER
void Control(int ctrl_id, vpx_active_map_t *arg) {
const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);

365
test/error_resilience_test.cc
Просмотреть файл

@ -20,11 +20,10 @@ const int kMaxErrorFrames = 12;
const int kMaxDroppableFrames = 12;
class ErrorResilienceTestLarge : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, bool> {
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
ErrorResilienceTestLarge()
: EncoderTest(GET_PARAM(0)),
svc_support_(GET_PARAM(2)),
psnr_(0.0),
nframes_(0),
mismatch_psnr_(0.0),
@ -58,77 +57,23 @@ class ErrorResilienceTestLarge : public ::libvpx_test::EncoderTest,
nframes_++;
}
//
// Frame flags and layer id for temporal layers.
// For two layers, test pattern is:
// 1 3
// 0 2 .....
// LAST is updated on base/layer 0, GOLDEN updated on layer 1.
// Non-zero pattern_switch parameter means pattern will switch to
// not using LAST for frame_num >= pattern_switch.
int SetFrameFlags(int frame_num,
int num_temp_layers,
int pattern_switch) {
int frame_flags = 0;
if (num_temp_layers == 2) {
if (frame_num % 2 == 0) {
if (frame_num < pattern_switch || pattern_switch == 0) {
// Layer 0: predict from LAST and ARF, update LAST.
frame_flags = VP8_EFLAG_NO_REF_GF |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
} else {
// Layer 0: predict from GF and ARF, update GF.
frame_flags = VP8_EFLAG_NO_REF_LAST |
VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
}
} else {
if (frame_num < pattern_switch || pattern_switch == 0) {
// Layer 1: predict from L, GF, and ARF, update GF.
frame_flags = VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_UPD_LAST;
} else {
// Layer 1: predict from GF and ARF, update GF.
frame_flags = VP8_EFLAG_NO_REF_LAST |
VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ARF;
}
}
}
return frame_flags;
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
::libvpx_test::Encoder * /*encoder*/) {
frame_flags_ &= ~(VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
// For temporal layer case.
if (cfg_.ts_number_layers > 1) {
frame_flags_ = SetFrameFlags(video->frame(),
cfg_.ts_number_layers,
pattern_switch_);
if (droppable_nframes_ > 0 &&
(cfg_.g_pass == VPX_RC_LAST_PASS || cfg_.g_pass == VPX_RC_ONE_PASS)) {
for (unsigned int i = 0; i < droppable_nframes_; ++i) {
if (droppable_frames_[i] == video->frame()) {
std::cout << "Encoding droppable frame: "
<< droppable_frames_[i] << "\n";
frame_flags_ |= (VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
return;
}
}
} else {
if (droppable_nframes_ > 0 &&
(cfg_.g_pass == VPX_RC_LAST_PASS || cfg_.g_pass == VPX_RC_ONE_PASS)) {
for (unsigned int i = 0; i < droppable_nframes_; ++i) {
if (droppable_frames_[i] == video->frame()) {
std::cout << "Encoding droppable frame: "
<< droppable_frames_[i] << "\n";
frame_flags_ |= (VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
return;
}
}
}
}
}
@ -195,8 +140,6 @@ class ErrorResilienceTestLarge : public ::libvpx_test::EncoderTest,
pattern_switch_ = frame_switch;
}
bool svc_support_;
private:
double psnr_;
unsigned int nframes_;
@ -301,297 +244,5 @@ TEST_P(ErrorResilienceTestLarge, DropFramesWithoutRecovery) {
#endif
}
// Check for successful decoding and no encoder/decoder mismatch
// if we lose (i.e., drop before decoding) the enhancement layer frames for a
// two layer temporal pattern. The base layer does not predict from the top
// layer, so successful decoding is expected.
TEST_P(ErrorResilienceTestLarge, 2LayersDropEnhancement) {
// This test doesn't run if SVC is not supported.
if (!svc_support_)
return;
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 500;
cfg_.g_lag_in_frames = 0;
cfg_.rc_end_usage = VPX_CBR;
// 2 Temporal layers, no spatial layers, CBR mode.
cfg_.ss_number_layers = 1;
cfg_.ts_number_layers = 2;
cfg_.ts_rate_decimator[0] = 2;
cfg_.ts_rate_decimator[1] = 1;
cfg_.ts_periodicity = 2;
cfg_.ts_target_bitrate[0] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.ts_target_bitrate[1] = cfg_.rc_target_bitrate;
init_flags_ = VPX_CODEC_USE_PSNR;
libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
timebase.den, timebase.num, 0, 40);
// Error resilient mode ON.
cfg_.g_error_resilient = 1;
cfg_.kf_mode = VPX_KF_DISABLED;
SetPatternSwitch(0);
// The odd frames are the enhancement layer for 2 layer pattern, so set
// those frames as droppable. Drop the last 7 frames.
unsigned int num_droppable_frames = 7;
unsigned int droppable_frame_list[] = {27, 29, 31, 33, 35, 37, 39};
SetDroppableFrames(num_droppable_frames, droppable_frame_list);
SetErrorFrames(num_droppable_frames, droppable_frame_list);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Test that no mismatches have been found
std::cout << " Mismatch frames: "
<< GetMismatchFrames() << "\n";
EXPECT_EQ(GetMismatchFrames(), (unsigned int) 0);
// Reset previously set of error/droppable frames.
Reset();
}
// Check for successful decoding and no encoder/decoder mismatch
// for a two layer temporal pattern, where at some point in the
// sequence, the LAST ref is not used anymore.
TEST_P(ErrorResilienceTestLarge, 2LayersNoRefLast) {
// This test doesn't run if SVC is not supported.
if (!svc_support_)
return;
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 500;
cfg_.g_lag_in_frames = 0;
cfg_.rc_end_usage = VPX_CBR;
// 2 Temporal layers, no spatial layers, CBR mode.
cfg_.ss_number_layers = 1;
cfg_.ts_number_layers = 2;
cfg_.ts_rate_decimator[0] = 2;
cfg_.ts_rate_decimator[1] = 1;
cfg_.ts_periodicity = 2;
cfg_.ts_target_bitrate[0] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.ts_target_bitrate[1] = cfg_.rc_target_bitrate;
init_flags_ = VPX_CODEC_USE_PSNR;
libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
timebase.den, timebase.num, 0, 100);
// Error resilient mode ON.
cfg_.g_error_resilient = 1;
cfg_.kf_mode = VPX_KF_DISABLED;
SetPatternSwitch(60);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Test that no mismatches have been found
std::cout << " Mismatch frames: "
<< GetMismatchFrames() << "\n";
EXPECT_EQ(GetMismatchFrames(), (unsigned int) 0);
// Reset previously set of error/droppable frames.
Reset();
}
class ErrorResilienceTestLargeCodecControls : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
ErrorResilienceTestLargeCodecControls()
: EncoderTest(GET_PARAM(0)),
encoding_mode_(GET_PARAM(1)) {
Reset();
}
virtual ~ErrorResilienceTestLargeCodecControls() {}
void Reset() {
last_pts_ = 0;
tot_frame_number_ = 0;
// For testing up to 3 layers.
for (int i = 0; i < 3; ++i) {
bits_total_[i] = 0;
}
duration_ = 0.0;
}
virtual void SetUp() {
InitializeConfig();
SetMode(encoding_mode_);
}
//
// Frame flags and layer id for temporal layers.
//
// For two layers, test pattern is:
// 1 3
// 0 2 .....
// For three layers, test pattern is:
// 1 3 5 7
// 2 6
// 0 4 ....
// LAST is always update on base/layer 0, GOLDEN is updated on layer 1,
// and ALTREF is updated on top layer for 3 layer pattern.
int SetFrameFlags(int frame_num, int num_temp_layers) {
int frame_flags = 0;
if (num_temp_layers == 2) {
if (frame_num % 2 == 0) {
// Layer 0: predict from L and ARF, update L.
frame_flags = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF;
} else {
// Layer 1: predict from L, G and ARF, and update G.
frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_ENTROPY;
}
} else if (num_temp_layers == 3) {
if (frame_num % 4 == 0) {
// Layer 0: predict from L, update L.
frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
} else if ((frame_num - 2) % 4 == 0) {
// Layer 1: predict from L, G, update G.
frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_REF_ARF;
} else if ((frame_num - 1) % 2 == 0) {
// Layer 2: predict from L, G, ARF; update ARG.
frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_LAST;
}
}
return frame_flags;
}
int SetLayerId(int frame_num, int num_temp_layers) {
int layer_id = 0;
if (num_temp_layers == 2) {
if (frame_num % 2 == 0) {
layer_id = 0;
} else {
layer_id = 1;
}
} else if (num_temp_layers == 3) {
if (frame_num % 4 == 0) {
layer_id = 0;
} else if ((frame_num - 2) % 4 == 0) {
layer_id = 1;
} else if ((frame_num - 1) % 2 == 0) {
layer_id = 2;
}
}
return layer_id;
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
libvpx_test::Encoder *encoder) {
if (cfg_.ts_number_layers > 1) {
int layer_id = SetLayerId(video->frame(), cfg_.ts_number_layers);
int frame_flags = SetFrameFlags(video->frame(), cfg_.ts_number_layers);
if (video->frame() > 0) {
encoder->Control(VP8E_SET_TEMPORAL_LAYER_ID, layer_id);
encoder->Control(VP8E_SET_FRAME_FLAGS, frame_flags);
}
const vpx_rational_t tb = video->timebase();
timebase_ = static_cast<double>(tb.num) / tb.den;
duration_ = 0;
return;
}
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
// Time since last timestamp = duration.
vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
if (duration > 1) {
// Update counter for total number of frames (#frames input to encoder).
// Needed for setting the proper layer_id below.
tot_frame_number_ += static_cast<int>(duration - 1);
}
int layer = SetLayerId(tot_frame_number_, cfg_.ts_number_layers);
const size_t frame_size_in_bits = pkt->data.frame.sz * 8;
// Update the total encoded bits. For temporal layers, update the cumulative
// encoded bits per layer.
for (int i = layer; i < static_cast<int>(cfg_.ts_number_layers); ++i) {
bits_total_[i] += frame_size_in_bits;
}
// Update the most recent pts.
last_pts_ = pkt->data.frame.pts;
++tot_frame_number_;
}
virtual void EndPassHook(void) {
duration_ = (last_pts_ + 1) * timebase_;
if (cfg_.ts_number_layers > 1) {
for (int layer = 0; layer < static_cast<int>(cfg_.ts_number_layers);
++layer) {
if (bits_total_[layer]) {
// Effective file datarate:
effective_datarate_[layer] = (bits_total_[layer] / 1000.0) / duration_;
}
}
}
}
double effective_datarate_[3];
private:
libvpx_test::TestMode encoding_mode_;
vpx_codec_pts_t last_pts_;
double timebase_;
int64_t bits_total_[3];
double duration_;
int tot_frame_number_;
};
// Check two codec controls used for:
// (1) for setting temporal layer id, and (2) for settings encoder flags.
// This test invokes those controls for each frame, and verifies encoder/decoder
// mismatch and basic rate control response.
// TODO(marpan): Maybe move this test to datarate_test.cc.
TEST_P(ErrorResilienceTestLargeCodecControls, CodecControl3TemporalLayers) {
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 500;
cfg_.rc_buf_sz = 1000;
cfg_.rc_dropframe_thresh = 1;
cfg_.rc_min_quantizer = 2;
cfg_.rc_max_quantizer = 56;
cfg_.rc_end_usage = VPX_CBR;
cfg_.rc_dropframe_thresh = 1;
cfg_.g_lag_in_frames = 0;
cfg_.kf_mode = VPX_KF_DISABLED;
cfg_.g_error_resilient = 1;
// 3 Temporal layers. Framerate decimation (4, 2, 1).
cfg_.ts_number_layers = 3;
cfg_.ts_rate_decimator[0] = 4;
cfg_.ts_rate_decimator[1] = 2;
cfg_.ts_rate_decimator[2] = 1;
cfg_.ts_periodicity = 4;
cfg_.ts_layer_id[0] = 0;
cfg_.ts_layer_id[1] = 2;
cfg_.ts_layer_id[2] = 1;
cfg_.ts_layer_id[3] = 2;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 200);
for (int i = 200; i <= 800; i += 200) {
cfg_.rc_target_bitrate = i;
Reset();
// 40-20-40 bitrate allocation for 3 temporal layers.
cfg_.ts_target_bitrate[0] = 40 * cfg_.rc_target_bitrate / 100;
cfg_.ts_target_bitrate[1] = 60 * cfg_.rc_target_bitrate / 100;
cfg_.ts_target_bitrate[2] = cfg_.rc_target_bitrate;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {
ASSERT_GE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 0.75)
<< " The datarate for the file is lower than target by too much, "
"for layer: " << j;
ASSERT_LE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 1.25)
<< " The datarate for the file is greater than target by too much, "
"for layer: " << j;
}
}
}
// SVC-related tests don't run for VP10 since SVC is not supported.
VP10_INSTANTIATE_TEST_CASE(ErrorResilienceTestLarge, ONE_PASS_TEST_MODES,
::testing::Values(false));
VP10_INSTANTIATE_TEST_CASE(ErrorResilienceTestLarge, ONE_PASS_TEST_MODES);
} // namespace

72
test/spatial_svc_encoder.sh
Просмотреть файл

@ -1,72 +0,0 @@
#!/bin/sh
##
## Copyright (c) 2014 The WebM project authors. All Rights Reserved.
##
## Use of this source code is governed by a BSD-style license
## that can be found in the LICENSE file in the root of the source
## tree. An additional intellectual property rights grant can be found
## in the file PATENTS. All contributing project authors may
## be found in the AUTHORS file in the root of the source tree.
##
## This file tests the libvpx vp9_spatial_svc_encoder example. To add new
## tests to to this file, do the following:
## 1. Write a shell function (this is your test).
## 2. Add the function to vp9_spatial_svc_tests (on a new line).
##
. $(dirname $0)/tools_common.sh
# Environment check: $YUV_RAW_INPUT is required.
vp9_spatial_svc_encoder_verify_environment() {
if [ ! -e "${YUV_RAW_INPUT}" ]; then
echo "Libvpx test data must exist in LIBVPX_TEST_DATA_PATH."
return 1
fi
}
# Runs vp9_spatial_svc_encoder. $1 is the test name.
vp9_spatial_svc_encoder() {
local readonly \
encoder="${LIBVPX_BIN_PATH}/vp9_spatial_svc_encoder${VPX_TEST_EXE_SUFFIX}"
local readonly test_name="$1"
local readonly \
output_file="${VPX_TEST_OUTPUT_DIR}/vp9_ssvc_encoder${test_name}.ivf"
local readonly frames_to_encode=10
local readonly max_kf=9999
shift
if [ ! -x "${encoder}" ]; then
elog "${encoder} does not exist or is not executable."
return 1
fi
eval "${VPX_TEST_PREFIX}" "${encoder}" -w "${YUV_RAW_INPUT_WIDTH}" \
-h "${YUV_RAW_INPUT_HEIGHT}" -k "${max_kf}" -f "${frames_to_encode}" \
"$@" "${YUV_RAW_INPUT}" "${output_file}" ${devnull}
[ -e "${output_file}" ] || return 1
}
# Each test is run with layer count 1-$vp9_ssvc_test_layers.
vp9_ssvc_test_layers=5
vp9_spatial_svc() {
if [ "$(vp9_encode_available)" = "yes" ]; then
local readonly test_name="vp9_spatial_svc"
for layers in $(seq 1 ${vp9_ssvc_test_layers}); do
vp9_spatial_svc_encoder "${test_name}" -sl ${layers}
done
fi
}
readonly vp9_spatial_svc_tests="DISABLED_vp9_spatial_svc_mode_i
DISABLED_vp9_spatial_svc_mode_altip
DISABLED_vp9_spatial_svc_mode_ip
DISABLED_vp9_spatial_svc_mode_gf
vp9_spatial_svc"
if [ "$(vpx_config_option_enabled CONFIG_SPATIAL_SVC)" = "yes" ]; then
run_tests \
vp9_spatial_svc_encoder_verify_environment \
"${vp9_spatial_svc_tests}"
fi

803
test/svc_test.cc
Просмотреть файл

@ -1,803 +0,0 @@
/*
* Copyright (c) 2013 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <string>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/decode_test_driver.h"
#include "test/i420_video_source.h"
#include "vp9/decoder/vp9_decoder.h"
#include "vpx/svc_context.h"
#include "vpx/vp8cx.h"
#include "vpx/vpx_encoder.h"
namespace {
using libvpx_test::CodecFactory;
using libvpx_test::Decoder;
using libvpx_test::DxDataIterator;
using libvpx_test::VP9CodecFactory;
class SvcTest : public ::testing::Test {
protected:
static const uint32_t kWidth = 352;
static const uint32_t kHeight = 288;
SvcTest()
: codec_iface_(0),
test_file_name_("hantro_collage_w352h288.yuv"),
codec_initialized_(false),
decoder_(0) {
memset(&svc_, 0, sizeof(svc_));
memset(&codec_, 0, sizeof(codec_));
memset(&codec_enc_, 0, sizeof(codec_enc_));
}
virtual ~SvcTest() {}
virtual void SetUp() {
svc_.log_level = SVC_LOG_DEBUG;
svc_.log_print = 0;
codec_iface_ = vpx_codec_vp9_cx();
const vpx_codec_err_t res =
vpx_codec_enc_config_default(codec_iface_, &codec_enc_, 0);
EXPECT_EQ(VPX_CODEC_OK, res);
codec_enc_.g_w = kWidth;
codec_enc_.g_h = kHeight;
codec_enc_.g_timebase.num = 1;
codec_enc_.g_timebase.den = 60;
codec_enc_.kf_min_dist = 100;
codec_enc_.kf_max_dist = 100;
vpx_codec_dec_cfg_t dec_cfg = vpx_codec_dec_cfg_t();
VP9CodecFactory codec_factory;
decoder_ = codec_factory.CreateDecoder(dec_cfg, 0);
#if CONFIG_VP10 && CONFIG_EXT_TILE
if (decoder_->IsVP10()) {
decoder_->Control(VP10_SET_DECODE_TILE_ROW, -1);
decoder_->Control(VP10_SET_DECODE_TILE_COL, -1);
}
#endif
tile_columns_ = 0;
tile_rows_ = 0;
}
virtual void TearDown() {
ReleaseEncoder();
delete(decoder_);
}
void InitializeEncoder() {
const vpx_codec_err_t res =
vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_OK, res);
vpx_codec_control(&codec_, VP8E_SET_CPUUSED, 4); // Make the test faster
vpx_codec_control(&codec_, VP9E_SET_TILE_COLUMNS, tile_columns_);
vpx_codec_control(&codec_, VP9E_SET_TILE_ROWS, tile_rows_);
codec_initialized_ = true;
}
void ReleaseEncoder() {
vpx_svc_release(&svc_);
if (codec_initialized_) vpx_codec_destroy(&codec_);
codec_initialized_ = false;
}
void GetStatsData(std::string *const stats_buf) {
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *cx_pkt;
while ((cx_pkt = vpx_codec_get_cx_data(&codec_, &iter)) != NULL) {
if (cx_pkt->kind == VPX_CODEC_STATS_PKT) {
EXPECT_GT(cx_pkt->data.twopass_stats.sz, 0U);
ASSERT_TRUE(cx_pkt->data.twopass_stats.buf != NULL);
stats_buf->append(static_cast<char*>(cx_pkt->data.twopass_stats.buf),
cx_pkt->data.twopass_stats.sz);
}
}
}
void Pass1EncodeNFrames(const int n, const int layers,
std::string *const stats_buf) {
vpx_codec_err_t res;
ASSERT_GT(n, 0);
ASSERT_GT(layers, 0);
svc_.spatial_layers = layers;
codec_enc_.g_pass = VPX_RC_FIRST_PASS;
InitializeEncoder();
libvpx_test::I420VideoSource video(test_file_name_,
codec_enc_.g_w, codec_enc_.g_h,
codec_enc_.g_timebase.den,
codec_enc_.g_timebase.num, 0, 30);
video.Begin();
for (int i = 0; i < n; ++i) {
res = vpx_svc_encode(&svc_, &codec_, video.img(), video.pts(),
video.duration(), VPX_DL_GOOD_QUALITY);
ASSERT_EQ(VPX_CODEC_OK, res);
GetStatsData(stats_buf);
video.Next();
}
// Flush encoder and test EOS packet.
res = vpx_svc_encode(&svc_, &codec_, NULL, video.pts(),
video.duration(), VPX_DL_GOOD_QUALITY);
ASSERT_EQ(VPX_CODEC_OK, res);
GetStatsData(stats_buf);
ReleaseEncoder();
}
void StoreFrames(const size_t max_frame_received,
struct vpx_fixed_buf *const outputs,
size_t *const frame_received) {
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *cx_pkt;
while ((cx_pkt = vpx_codec_get_cx_data(&codec_, &iter)) != NULL) {
if (cx_pkt->kind == VPX_CODEC_CX_FRAME_PKT) {
const size_t frame_size = cx_pkt->data.frame.sz;
EXPECT_GT(frame_size, 0U);
ASSERT_TRUE(cx_pkt->data.frame.buf != NULL);
ASSERT_LT(*frame_received, max_frame_received);
if (*frame_received == 0)
EXPECT_EQ(1, !!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY));
outputs[*frame_received].buf = malloc(frame_size + 16);
ASSERT_TRUE(outputs[*frame_received].buf != NULL);
memcpy(outputs[*frame_received].buf, cx_pkt->data.frame.buf,
frame_size);
outputs[*frame_received].sz = frame_size;
++(*frame_received);
}
}
}
void Pass2EncodeNFrames(std::string *const stats_buf,
const int n, const int layers,
struct vpx_fixed_buf *const outputs) {
vpx_codec_err_t res;
size_t frame_received = 0;
ASSERT_TRUE(outputs != NULL);
ASSERT_GT(n, 0);
ASSERT_GT(layers, 0);
svc_.spatial_layers = layers;
codec_enc_.rc_target_bitrate = 500;
if (codec_enc_.g_pass == VPX_RC_LAST_PASS) {
ASSERT_TRUE(stats_buf != NULL);
ASSERT_GT(stats_buf->size(), 0U);
codec_enc_.rc_twopass_stats_in.buf = &(*stats_buf)[0];
codec_enc_.rc_twopass_stats_in.sz = stats_buf->size();
}
InitializeEncoder();
libvpx_test::I420VideoSource video(test_file_name_,
codec_enc_.g_w, codec_enc_.g_h,
codec_enc_.g_timebase.den,
codec_enc_.g_timebase.num, 0, 30);
video.Begin();
for (int i = 0; i < n; ++i) {
res = vpx_svc_encode(&svc_, &codec_, video.img(), video.pts(),
video.duration(), VPX_DL_GOOD_QUALITY);
ASSERT_EQ(VPX_CODEC_OK, res);
StoreFrames(n, outputs, &frame_received);
video.Next();
}
// Flush encoder.
res = vpx_svc_encode(&svc_, &codec_, NULL, 0,
video.duration(), VPX_DL_GOOD_QUALITY);
EXPECT_EQ(VPX_CODEC_OK, res);
StoreFrames(n, outputs, &frame_received);
EXPECT_EQ(frame_received, static_cast<size_t>(n));
ReleaseEncoder();
}
void DecodeNFrames(const struct vpx_fixed_buf *const inputs, const int n) {
int decoded_frames = 0;
int received_frames = 0;
ASSERT_TRUE(inputs != NULL);
ASSERT_GT(n, 0);
for (int i = 0; i < n; ++i) {
ASSERT_TRUE(inputs[i].buf != NULL);
ASSERT_GT(inputs[i].sz, 0U);
const vpx_codec_err_t res_dec =
decoder_->DecodeFrame(static_cast<const uint8_t *>(inputs[i].buf),
inputs[i].sz);
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder_->DecodeError();
++decoded_frames;
DxDataIterator dec_iter = decoder_->GetDxData();
while (dec_iter.Next() != NULL) {
++received_frames;
}
}
EXPECT_EQ(decoded_frames, n);
EXPECT_EQ(received_frames, n);
}
void DropEnhancementLayers(struct vpx_fixed_buf *const inputs,
const int num_super_frames,
const int remained_spatial_layers) {
ASSERT_TRUE(inputs != NULL);
ASSERT_GT(num_super_frames, 0);
ASSERT_GT(remained_spatial_layers, 0);
for (int i = 0; i < num_super_frames; ++i) {
uint32_t frame_sizes[8] = {0};
int frame_count = 0;
int frames_found = 0;
int frame;
ASSERT_TRUE(inputs[i].buf != NULL);
ASSERT_GT(inputs[i].sz, 0U);
vpx_codec_err_t res =
vp9_parse_superframe_index(static_cast<const uint8_t*>(inputs[i].buf),
inputs[i].sz, frame_sizes, &frame_count,
NULL, NULL);
ASSERT_EQ(VPX_CODEC_OK, res);
if (frame_count == 0) {
// There's no super frame but only a single frame.
ASSERT_EQ(1, remained_spatial_layers);
} else {
// Found a super frame.
uint8_t *frame_data = static_cast<uint8_t*>(inputs[i].buf);
uint8_t *frame_start = frame_data;
for (frame = 0; frame < frame_count; ++frame) {
// Looking for a visible frame.
if (frame_data[0] & 0x02) {
++frames_found;
if (frames_found == remained_spatial_layers)
break;
}
frame_data += frame_sizes[frame];
}
ASSERT_LT(frame, frame_count) << "Couldn't find a visible frame. "
<< "remained_spatial_layers: " << remained_spatial_layers
<< " super_frame: " << i;
if (frame == frame_count - 1)
continue;
frame_data += frame_sizes[frame];
// We need to add one more frame for multiple frame contexts.
uint8_t marker =
static_cast<const uint8_t*>(inputs[i].buf)[inputs[i].sz - 1];
const uint32_t mag = ((marker >> 3) & 0x3) + 1;
const size_t index_sz = 2 + mag * frame_count;
const size_t new_index_sz = 2 + mag * (frame + 1);
marker &= 0x0f8;
marker |= frame;
// Copy existing frame sizes.
memmove(frame_data + 1, frame_start + inputs[i].sz - index_sz + 1,
new_index_sz - 2);
// New marker.
frame_data[0] = marker;
frame_data += (mag * (frame + 1) + 1);
*frame_data++ = marker;
inputs[i].sz = frame_data - frame_start;
}
}
}
void FreeBitstreamBuffers(struct vpx_fixed_buf *const inputs, const int n) {
ASSERT_TRUE(inputs != NULL);
ASSERT_GT(n, 0);
for (int i = 0; i < n; ++i) {
free(inputs[i].buf);
inputs[i].buf = NULL;
inputs[i].sz = 0;
}
}
SvcContext svc_;
vpx_codec_ctx_t codec_;
struct vpx_codec_enc_cfg codec_enc_;
vpx_codec_iface_t *codec_iface_;
std::string test_file_name_;
bool codec_initialized_;
Decoder *decoder_;
int tile_columns_;
int tile_rows_;
};
TEST_F(SvcTest, SvcInit) {
// test missing parameters
vpx_codec_err_t res = vpx_svc_init(NULL, &codec_, codec_iface_, &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_init(&svc_, NULL, codec_iface_, &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_init(&svc_, &codec_, NULL, &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_init(&svc_, &codec_, codec_iface_, NULL);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
svc_.spatial_layers = 6; // too many layers
res = vpx_svc_init(&svc_, &codec_, codec_iface_, &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
svc_.spatial_layers = 0; // use default layers
InitializeEncoder();
EXPECT_EQ(VPX_SS_DEFAULT_LAYERS, svc_.spatial_layers);
}
TEST_F(SvcTest, InitTwoLayers) {
svc_.spatial_layers = 2;
InitializeEncoder();
}
TEST_F(SvcTest, InvalidOptions) {
vpx_codec_err_t res = vpx_svc_set_options(&svc_, NULL);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "not-an-option=1");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
}
TEST_F(SvcTest, SetLayersOption) {
vpx_codec_err_t res = vpx_svc_set_options(&svc_, "spatial-layers=3");
EXPECT_EQ(VPX_CODEC_OK, res);
InitializeEncoder();
EXPECT_EQ(3, svc_.spatial_layers);
}
TEST_F(SvcTest, SetMultipleOptions) {
vpx_codec_err_t res =
vpx_svc_set_options(&svc_, "spatial-layers=2 scale-factors=1/3,2/3");
EXPECT_EQ(VPX_CODEC_OK, res);
InitializeEncoder();
EXPECT_EQ(2, svc_.spatial_layers);
}
TEST_F(SvcTest, SetScaleFactorsOption) {
svc_.spatial_layers = 2;
vpx_codec_err_t res =
vpx_svc_set_options(&svc_, "scale-factors=not-scale-factors");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "scale-factors=1/3, 3*3");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "scale-factors=1/3");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "scale-factors=1/3,2/3");
EXPECT_EQ(VPX_CODEC_OK, res);
InitializeEncoder();
}
TEST_F(SvcTest, SetQuantizersOption) {
svc_.spatial_layers = 2;
vpx_codec_err_t res = vpx_svc_set_options(&svc_, "max-quantizers=nothing");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "min-quantizers=nothing");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "max-quantizers=40");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "min-quantizers=40");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "max-quantizers=30,30 min-quantizers=40,40");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "max-quantizers=40,40 min-quantizers=30,30");
InitializeEncoder();
}
TEST_F(SvcTest, SetAutoAltRefOption) {
svc_.spatial_layers = 5;
vpx_codec_err_t res = vpx_svc_set_options(&svc_, "auto-alt-refs=none");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
res = vpx_svc_set_options(&svc_, "auto-alt-refs=1,1,1,1,0");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
vpx_svc_set_options(&svc_, "auto-alt-refs=0,1,1,1,0");
InitializeEncoder();
}
// Test that decoder can handle an SVC frame as the first frame in a sequence.
TEST_F(SvcTest, OnePassEncodeOneFrame) {
codec_enc_.g_pass = VPX_RC_ONE_PASS;
vpx_fixed_buf output = {0};
Pass2EncodeNFrames(NULL, 1, 2, &output);
DecodeNFrames(&output, 1);
FreeBitstreamBuffers(&output, 1);
}
TEST_F(SvcTest, OnePassEncodeThreeFrames) {
codec_enc_.g_pass = VPX_RC_ONE_PASS;
codec_enc_.g_lag_in_frames = 0;
vpx_fixed_buf outputs[3];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(NULL, 3, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 3);
FreeBitstreamBuffers(&outputs[0], 3);
}
TEST_F(SvcTest, TwoPassEncode10Frames) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(10, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode20FramesWithAltRef) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(20, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1");
vpx_fixed_buf outputs[20];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 20, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 20);
FreeBitstreamBuffers(&outputs[0], 20);
}
TEST_F(SvcTest, TwoPassEncode2SpatialLayersDecodeBaseLayerOnly) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(10, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode5SpatialLayersDecode54321Layers) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(10, 5, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_svc_set_options(&svc_, "auto-alt-refs=0,1,1,1,0");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 5, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 4);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 3);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 2);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2SNRLayers) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1,1/1");
Pass1EncodeNFrames(20, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_svc_set_options(&svc_,
"auto-alt-refs=1,1 scale-factors=1/1,1/1");
vpx_fixed_buf outputs[20];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 20, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 20);
FreeBitstreamBuffers(&outputs[0], 20);
}
TEST_F(SvcTest, TwoPassEncode3SNRLayersDecode321Layers) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1,1/1,1/1");
Pass1EncodeNFrames(20, 3, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
vpx_svc_set_options(&svc_,
"auto-alt-refs=1,1,1 scale-factors=1/1,1/1,1/1");
vpx_fixed_buf outputs[20];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 20, 3, &outputs[0]);
DecodeNFrames(&outputs[0], 20);
DropEnhancementLayers(&outputs[0], 20, 2);
DecodeNFrames(&outputs[0], 20);
DropEnhancementLayers(&outputs[0], 20, 1);
DecodeNFrames(&outputs[0], 20);
FreeBitstreamBuffers(&outputs[0], 20);
}
TEST_F(SvcTest, SetMultipleFrameContextsOption) {
svc_.spatial_layers = 5;
vpx_codec_err_t res =
vpx_svc_set_options(&svc_, "multi-frame-contexts=1");
EXPECT_EQ(VPX_CODEC_OK, res);
res = vpx_svc_init(&svc_, &codec_, vpx_codec_vp9_cx(), &codec_enc_);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
svc_.spatial_layers = 2;
res = vpx_svc_set_options(&svc_, "multi-frame-contexts=1");
InitializeEncoder();
}
TEST_F(SvcTest, TwoPassEncode2SpatialLayersWithMultipleFrameContexts) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(10, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1 multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest,
TwoPassEncode2SpatialLayersWithMultipleFrameContextsDecodeBaselayer) {
// First pass encode
std::string stats_buf;
Pass1EncodeNFrames(10, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1 multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2SNRLayersWithMultipleFrameContexts) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1,1/1");
Pass1EncodeNFrames(10, 2, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1 scale-factors=1/1,1/1 "
"multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest,
TwoPassEncode3SNRLayersWithMultipleFrameContextsDecode321Layer) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1,1/1,1/1");
Pass1EncodeNFrames(10, 3, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1,1,1 scale-factors=1/1,1/1,1/1 "
"multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 3, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 2);
DecodeNFrames(&outputs[0], 10);
DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayers) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayersWithMultipleFrameContexts) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1 "
"multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayersDecodeBaseLayer) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
vpx_fixed_buf base_layer[5];
for (int i = 0; i < 5; ++i)
base_layer[i] = outputs[i * 2];
DecodeNFrames(&base_layer[0], 5);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest,
TwoPassEncode2TemporalLayersWithMultipleFrameContextsDecodeBaseLayer) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
codec_enc_.g_error_resilient = 0;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1 "
"multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
vpx_fixed_buf base_layer[5];
for (int i = 0; i < 5; ++i)
base_layer[i] = outputs[i * 2];
DecodeNFrames(&base_layer[0], 5);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayersWithTiles) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1");
codec_enc_.g_w = 704;
codec_enc_.g_h = 144;
tile_columns_ = 1;
tile_rows_ = 1;
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
TEST_F(SvcTest,
TwoPassEncode2TemporalLayersWithMultipleFrameContextsAndTiles) {
// First pass encode
std::string stats_buf;
vpx_svc_set_options(&svc_, "scale-factors=1/1");
svc_.temporal_layers = 2;
Pass1EncodeNFrames(10, 1, &stats_buf);
// Second pass encode
codec_enc_.g_pass = VPX_RC_LAST_PASS;
svc_.temporal_layers = 2;
codec_enc_.g_error_resilient = 0;
codec_enc_.g_w = 704;
codec_enc_.g_h = 144;
tile_columns_ = 1;
tile_rows_ = 1;
vpx_svc_set_options(&svc_, "auto-alt-refs=1 scale-factors=1/1 "
"multi-frame-contexts=1");
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
} // namespace

4
test/test.mk
Просмотреть файл

@ -120,10 +120,6 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += error_block_test.cc
#LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += vp9_quantize_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += subtract_test.cc
ifeq ($(CONFIG_VP10_ENCODER),yes)
LIBVPX_TEST_SRCS-$(CONFIG_SPATIAL_SVC) += svc_test.cc
endif
ifeq ($(CONFIG_VP10_ENCODER)$(CONFIG_VP10_TEMPORAL_DENOISING),yesyes)
LIBVPX_TEST_SRCS-$(HAVE_SSE2) += denoiser_sse2_test.cc
endif

290
test/vpx_temporal_svc_encoder.sh
Просмотреть файл

@ -1,290 +0,0 @@
#!/bin/sh
##
## Copyright (c) 2014 The WebM project authors. All Rights Reserved.
##
## Use of this source code is governed by a BSD-style license
## that can be found in the LICENSE file in the root of the source
## tree. An additional intellectual property rights grant can be found
## in the file PATENTS. All contributing project authors may
## be found in the AUTHORS file in the root of the source tree.
##
## This file tests the libvpx vpx_temporal_svc_encoder example. To add new
## tests to this file, do the following:
## 1. Write a shell function (this is your test).
## 2. Add the function to vpx_tsvc_encoder_tests (on a new line).
##
. $(dirname $0)/tools_common.sh
# Environment check: $YUV_RAW_INPUT is required.
vpx_tsvc_encoder_verify_environment() {
if [ ! -e "${YUV_RAW_INPUT}" ]; then
echo "Libvpx test data must exist in LIBVPX_TEST_DATA_PATH."
return 1
fi
if [ "$(vpx_config_option_enabled CONFIG_TEMPORAL_DENOISING)" != "yes" ]; then
elog "Warning: Temporal denoising is disabled! Spatial denoising will be " \
"used instead, which is probably not what you want for this test."
fi
}
# Runs vpx_temporal_svc_encoder using the codec specified by $1 and output file
# name by $2. Additional positional parameters are passed directly to
# vpx_temporal_svc_encoder.
vpx_tsvc_encoder() {
local encoder="${LIBVPX_BIN_PATH}/vpx_temporal_svc_encoder"
encoder="${encoder}${VPX_TEST_EXE_SUFFIX}"
local codec="$1"
local output_file_base="$2"
local output_file="${VPX_TEST_OUTPUT_DIR}/${output_file_base}"
local timebase_num="1"
local timebase_den="1000"
local speed="6"
local frame_drop_thresh="30"
shift 2
if [ ! -x "${encoder}" ]; then
elog "${encoder} does not exist or is not executable."
return 1
fi
eval "${VPX_TEST_PREFIX}" "${encoder}" "${YUV_RAW_INPUT}" "${output_file}" \
"${codec}" "${YUV_RAW_INPUT_WIDTH}" "${YUV_RAW_INPUT_HEIGHT}" \
"${timebase_num}" "${timebase_den}" "${speed}" "${frame_drop_thresh}" \
"$@" \
${devnull}
}
# Confirms that all expected output files exist given the output file name
# passed to vpx_temporal_svc_encoder.
# The file name passed to vpx_temporal_svc_encoder is joined with the stream
# number and the extension .ivf to produce per stream output files. Here $1 is
# file name, and $2 is expected number of files.
files_exist() {
local file_name="${VPX_TEST_OUTPUT_DIR}/$1"
local num_files="$(($2 - 1))"
for stream_num in $(seq 0 ${num_files}); do
[ -e "${file_name}_${stream_num}.ivf" ] || return 1
done
}
# Run vpx_temporal_svc_encoder in all supported modes for vp8 and vp9.
vpx_tsvc_encoder_vp8_mode_0() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 0 200 || return 1
# Mode 0 produces 1 stream
files_exist "${FUNCNAME}" 1 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_1() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 1 200 400 || return 1
# Mode 1 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_2() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 2 200 400 || return 1
# Mode 2 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_3() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 3 200 400 600 || return 1
# Mode 3 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_4() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 4 200 400 600 || return 1
# Mode 4 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_5() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 5 200 400 600 || return 1
# Mode 5 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_6() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 6 200 400 600 || return 1
# Mode 6 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_7() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 7 200 400 600 800 1000 || return 1
# Mode 7 produces 5 streams
files_exist "${FUNCNAME}" 5 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_8() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 8 200 400 || return 1
# Mode 8 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_9() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 9 200 400 600 || return 1
# Mode 9 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_10() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 10 200 400 600 || return 1
# Mode 10 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_11() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 11 200 400 600 || return 1
# Mode 11 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_0() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 0 200 || return 1
# Mode 0 produces 1 stream
files_exist "${FUNCNAME}" 1 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_1() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 1 200 400 || return 1
# Mode 1 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_2() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 2 200 400 || return 1
# Mode 2 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_3() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 3 200 400 600 || return 1
# Mode 3 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_4() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 4 200 400 600 || return 1
# Mode 4 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_5() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 5 200 400 600 || return 1
# Mode 5 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_6() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 6 200 400 600 || return 1
# Mode 6 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_7() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 7 200 400 600 800 1000 || return 1
# Mode 7 produces 5 streams
files_exist "${FUNCNAME}" 5 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_8() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 8 200 400 || return 1
# Mode 8 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_9() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 9 200 400 600 || return 1
# Mode 9 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_10() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 10 200 400 600 || return 1
# Mode 10 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_11() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 11 200 400 600 || return 1
# Mode 11 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
fi
}
vpx_tsvc_encoder_tests="vpx_tsvc_encoder_vp8_mode_0
vpx_tsvc_encoder_vp8_mode_1
vpx_tsvc_encoder_vp8_mode_2
vpx_tsvc_encoder_vp8_mode_3
vpx_tsvc_encoder_vp8_mode_4
vpx_tsvc_encoder_vp8_mode_5
vpx_tsvc_encoder_vp8_mode_6
vpx_tsvc_encoder_vp8_mode_7
vpx_tsvc_encoder_vp8_mode_8
vpx_tsvc_encoder_vp8_mode_9
vpx_tsvc_encoder_vp8_mode_10
vpx_tsvc_encoder_vp8_mode_11
vpx_tsvc_encoder_vp9_mode_0
vpx_tsvc_encoder_vp9_mode_1
vpx_tsvc_encoder_vp9_mode_2
vpx_tsvc_encoder_vp9_mode_3
vpx_tsvc_encoder_vp9_mode_4
vpx_tsvc_encoder_vp9_mode_5
vpx_tsvc_encoder_vp9_mode_6
vpx_tsvc_encoder_vp9_mode_7
vpx_tsvc_encoder_vp9_mode_8
vpx_tsvc_encoder_vp9_mode_9
vpx_tsvc_encoder_vp9_mode_10
vpx_tsvc_encoder_vp9_mode_11"
run_tests vpx_tsvc_encoder_verify_environment "${vpx_tsvc_encoder_tests}"

1
vp10/encoder/encoder.c
Просмотреть файл

@ -43,7 +43,6 @@
#include "vp10/encoder/rd.h"
#include "vp10/encoder/resize.h"
#include "vp10/encoder/segmentation.h"
#include "vp10/encoder/skin_detection.h"
#include "vp10/encoder/speed_features.h"
#include "vp10/encoder/temporal_filter.h"

104
vp10/encoder/skin_detection.c
Просмотреть файл

@ -1,104 +0,0 @@
/*
* Copyright (c) 2015 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <limits.h>
#include <math.h>
#include "vp10/common/blockd.h"
#include "vp10/encoder/encoder.h"
#include "vp10/encoder/skin_detection.h"
// Fixed-point skin color model parameters.
static const int skin_mean[2] = {7463, 9614}; // q6
static const int skin_inv_cov[4] = {4107, 1663, 1663, 2157}; // q16
static const int skin_threshold = 1570636; // q18
// Thresholds on luminance.
static const int y_low = 20;
static const int y_high = 220;
// Evaluates the Mahalanobis distance measure for the input CbCr values.
static int evaluate_skin_color_difference(int cb, int cr) {
const int cb_q6 = cb << 6;
const int cr_q6 = cr << 6;
const int cb_diff_q12 = (cb_q6 - skin_mean[0]) * (cb_q6 - skin_mean[0]);
const int cbcr_diff_q12 = (cb_q6 - skin_mean[0]) * (cr_q6 - skin_mean[1]);
const int cr_diff_q12 = (cr_q6 - skin_mean[1]) * (cr_q6 - skin_mean[1]);
const int cb_diff_q2 = (cb_diff_q12 + (1 << 9)) >> 10;
const int cbcr_diff_q2 = (cbcr_diff_q12 + (1 << 9)) >> 10;
const int cr_diff_q2 = (cr_diff_q12 + (1 << 9)) >> 10;
const int skin_diff = skin_inv_cov[0] * cb_diff_q2 +
skin_inv_cov[1] * cbcr_diff_q2 +
skin_inv_cov[2] * cbcr_diff_q2 +
skin_inv_cov[3] * cr_diff_q2;
return skin_diff;
}
int vp10_skin_pixel(const uint8_t y, const uint8_t cb, const uint8_t cr) {
if (y < y_low || y > y_high)
return 0;
else
return (evaluate_skin_color_difference(cb, cr) < skin_threshold);
}
#ifdef OUTPUT_YUV_SKINMAP
// For viewing skin map on input source.
void vp10_compute_skin_map(VP10_COMP *const cpi, FILE *yuv_skinmap_file) {
int i, j, mi_row, mi_col;
VP10_COMMON *const cm = &cpi->common;
uint8_t *y;
const uint8_t *src_y = cpi->Source->y_buffer;
const uint8_t *src_u = cpi->Source->u_buffer;
const uint8_t *src_v = cpi->Source->v_buffer;
const int src_ystride = cpi->Source->y_stride;
const int src_uvstride = cpi->Source->uv_stride;
YV12_BUFFER_CONFIG skinmap;
memset(&skinmap, 0, sizeof(YV12_BUFFER_CONFIG));
if (vpx_alloc_frame_buffer(&skinmap, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment)) {
vpx_free_frame_buffer(&skinmap);
return;
}
memset(skinmap.buffer_alloc, 128, skinmap.frame_size);
y = skinmap.y_buffer;
// Loop through 8x8 blocks and set skin map based on center pixel of block.
// Set y to white for skin block, otherwise set to source with gray scale.
// Ignore rightmost/bottom boundary blocks.
for (mi_row = 0; mi_row < cm->mi_rows - 1; ++mi_row) {
for (mi_col = 0; mi_col < cm->mi_cols - 1; ++mi_col) {
// Use middle pixel for each 8x8 block for skin detection.
// If middle pixel is skin, assign whole 8x8 block to skin.
const uint8_t ysource = src_y[4 * src_ystride + 4];
const uint8_t usource = src_u[2 * src_uvstride + 2];
const uint8_t vsource = src_v[2 * src_uvstride + 2];
const int is_skin = vp10_skin_pixel(ysource, usource, vsource);
for (i = 0; i < 8; i++) {
for (j = 0; j < 8; j++) {
if (is_skin)
y[i * src_ystride + j] = 255;
else
y[i * src_ystride + j] = src_y[i * src_ystride + j];
}
}
y += 8;
src_y += 8;
src_u += 4;
src_v += 4;
}
y += (src_ystride << 3) - ((cm->mi_cols - 1) << 3);
src_y += (src_ystride << 3) - ((cm->mi_cols - 1) << 3);
src_u += (src_uvstride << 2) - ((cm->mi_cols - 1) << 2);
src_v += (src_uvstride << 2) - ((cm->mi_cols - 1) << 2);
}
vp10_write_yuv_frame_420(&skinmap, yuv_skinmap_file);
vpx_free_frame_buffer(&skinmap);
}
#endif

35
vp10/encoder/skin_detection.h
Просмотреть файл

@ -1,35 +0,0 @@
/*
* Copyright (c) 2015 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP10_ENCODER_SKIN_MAP_H_
#define VP10_ENCODER_SKIN_MAP_H_
#include "vp10/common/blockd.h"
#ifdef __cplusplus
extern "C" {
#endif
struct VP10_COMP;
// #define OUTPUT_YUV_SKINMAP
int vp10_skin_pixel(const uint8_t y, const uint8_t cb, const uint8_t cr);
#ifdef OUTPUT_YUV_SKINMAP
// For viewing skin map on input source.
void vp10_compute_skin_map(VP10_COMP *const cpi, FILE *yuv_skinmap_file);
#endif
#ifdef __cplusplus
} // extern "C"
#endif
#endif // VP10_ENCODER_SKIN_MAP_H_

56
vp10/vp10_cx_iface.c
Просмотреть файл

@ -110,7 +110,6 @@ struct vpx_codec_alg_priv {
vp8_postproc_cfg_t preview_ppcfg;
vpx_codec_pkt_list_decl(256) pkt_list;
unsigned int fixed_kf_cntr;
vpx_codec_priv_output_cx_pkt_cb_pair_t output_cx_pkt_cb;
// BufferPool that holds all reference frames.
BufferPool *buffer_pool;
};
@ -192,10 +191,6 @@ static vpx_codec_err_t validate_config(vpx_codec_alg_priv_t *ctx,
RANGE_CHECK(cfg, rc_scaled_height, 0, cfg->g_h);
}
// Spatial/temporal scalability are not yet supported in VP10.
// Only accept the default value for range checking.
RANGE_CHECK(cfg, ss_number_layers, 1, 1);
RANGE_CHECK(cfg, ts_number_layers, 1, 1);
// VP9 does not support a lower bound on the keyframe interval in
// automatic keyframe placement mode.
if (cfg->kf_mode != VPX_KF_DISABLED &&
@ -1054,22 +1049,6 @@ static vpx_codec_err_t encoder_encode(vpx_codec_alg_priv_t *ctx,
cx_data += size;
cx_data_sz -= size;
if (ctx->output_cx_pkt_cb.output_cx_pkt) {
pkt.kind = VPX_CODEC_CX_FRAME_PKT;
pkt.data.frame.pts = ticks_to_timebase_units(timebase,
dst_time_stamp);
pkt.data.frame.duration =
(unsigned long)ticks_to_timebase_units(timebase,
dst_end_time_stamp - dst_time_stamp);
pkt.data.frame.flags = get_frame_pkt_flags(cpi, lib_flags);
pkt.data.frame.buf = ctx->pending_cx_data;
pkt.data.frame.sz = size;
ctx->pending_cx_data = NULL;
ctx->pending_cx_data_sz = 0;
ctx->pending_frame_count = 0;
ctx->output_cx_pkt_cb.output_cx_pkt(
&pkt, ctx->output_cx_pkt_cb.user_priv);
}
continue;
}
@ -1084,9 +1063,7 @@ static vpx_codec_err_t encoder_encode(vpx_codec_alg_priv_t *ctx,
if (ctx->pending_cx_data) {
ctx->pending_frame_sizes[ctx->pending_frame_count++] = size;
ctx->pending_cx_data_sz += size;
// write the superframe only for the case when
if (!ctx->output_cx_pkt_cb.output_cx_pkt)
size += write_superframe_index(ctx);
size += write_superframe_index(ctx);
pkt.data.frame.buf = ctx->pending_cx_data;
pkt.data.frame.sz = ctx->pending_cx_data_sz;
ctx->pending_cx_data = NULL;
@ -1098,11 +1075,7 @@ static vpx_codec_err_t encoder_encode(vpx_codec_alg_priv_t *ctx,
}
pkt.data.frame.partition_id = -1;
if(ctx->output_cx_pkt_cb.output_cx_pkt)
ctx->output_cx_pkt_cb.output_cx_pkt(&pkt,
ctx->output_cx_pkt_cb.user_priv);
else
vpx_codec_pkt_list_add(&ctx->pkt_list.head, &pkt);
vpx_codec_pkt_list_add(&ctx->pkt_list.head, &pkt);
cx_data += size;
cx_data_sz -= size;
@ -1265,16 +1238,6 @@ static vpx_codec_err_t ctrl_set_scale_mode(vpx_codec_alg_priv_t *ctx,
}
}
static vpx_codec_err_t ctrl_register_cx_callback(vpx_codec_alg_priv_t *ctx,
va_list args) {
vpx_codec_priv_output_cx_pkt_cb_pair_t *cbp =
(vpx_codec_priv_output_cx_pkt_cb_pair_t *)va_arg(args, void *);
ctx->output_cx_pkt_cb.output_cx_pkt = cbp->output_cx_pkt;
ctx->output_cx_pkt_cb.user_priv = cbp->user_priv;
return VPX_CODEC_OK;
}
static vpx_codec_err_t ctrl_set_tune_content(vpx_codec_alg_priv_t *ctx,
va_list args) {
struct vp10_extracfg extra_cfg = ctx->extra_cfg;
@ -1343,7 +1306,6 @@ static vpx_codec_ctrl_fn_map_t encoder_ctrl_maps[] = {
{VP9E_SET_FRAME_PARALLEL_DECODING, ctrl_set_frame_parallel_decoding_mode},
{VP9E_SET_AQ_MODE, ctrl_set_aq_mode},
{VP9E_SET_FRAME_PERIODIC_BOOST, ctrl_set_frame_periodic_boost},
{VP9E_REGISTER_CX_CALLBACK, ctrl_register_cx_callback},
{VP9E_SET_TUNE_CONTENT, ctrl_set_tune_content},
{VP9E_SET_COLOR_SPACE, ctrl_set_color_space},
{VP9E_SET_COLOR_RANGE, ctrl_set_color_range},
@ -1412,20 +1374,6 @@ static vpx_codec_enc_cfg_map_t encoder_usage_cfg_map[] = {
VPX_KF_AUTO, // g_kfmode
0, // kf_min_dist
9999, // kf_max_dist
// TODO(yunqingwang): Spatial/temporal scalability are not supported
// in VP10. The following 10 parameters are not used, which should
// be removed later.
1, // ss_number_layers
{0},
{0}, // ss_target_bitrate
1, // ts_number_layers
{0}, // ts_target_bitrate
{0}, // ts_rate_decimator
0, // ts_periodicity
{0}, // ts_layer_id
{0}, // layer_taget_bitrate
0 // temporal_layering_mode
}
},
};

2
vp10/vp10cx.mk
Просмотреть файл

@ -84,8 +84,6 @@ VP10_CX_SRCS-yes += encoder/aq_cyclicrefresh.c
VP10_CX_SRCS-yes += encoder/aq_cyclicrefresh.h
VP10_CX_SRCS-yes += encoder/aq_complexity.c
VP10_CX_SRCS-yes += encoder/aq_complexity.h
VP10_CX_SRCS-yes += encoder/skin_detection.c
VP10_CX_SRCS-yes += encoder/skin_detection.h
VP10_CX_SRCS-yes += encoder/temporal_filter.c
VP10_CX_SRCS-yes += encoder/temporal_filter.h
VP10_CX_SRCS-yes += encoder/mbgraph.c

6
vpx/exports_spatial_svc
Просмотреть файл

@ -1,6 +0,0 @@
text vpx_svc_dump_statistics
text vpx_svc_encode
text vpx_svc_get_message
text vpx_svc_init
text vpx_svc_release
text vpx_svc_set_options

686
vpx/src/svc_encodeframe.c
Просмотреть файл

@ -1,686 +0,0 @@
/*
* Copyright (c) 2013 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
/**
* @file
* VP9 SVC encoding support via libvpx
*/
#include <assert.h>
#include <math.h>
#include <limits.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#define VPX_DISABLE_CTRL_TYPECHECKS 1
#include "./vpx_config.h"
#include "vpx/svc_context.h"
#include "vpx/vp8cx.h"
#include "vpx/vpx_encoder.h"
#include "vpx_mem/vpx_mem.h"
#include "vp9/common/vp9_onyxc_int.h"
#ifdef __MINGW32__
#define strtok_r strtok_s
#ifndef MINGW_HAS_SECURE_API
// proto from /usr/x86_64-w64-mingw32/include/sec_api/string_s.h
_CRTIMP char *__cdecl strtok_s(char *str, const char *delim, char **context);
#endif /* MINGW_HAS_SECURE_API */
#endif /* __MINGW32__ */
#ifdef _MSC_VER
#define strdup _strdup
#define strtok_r strtok_s
#endif
#define SVC_REFERENCE_FRAMES 8
#define SUPERFRAME_SLOTS (8)
#define SUPERFRAME_BUFFER_SIZE (SUPERFRAME_SLOTS * sizeof(uint32_t) + 2)
#define MAX_QUANTIZER 63
static const int DEFAULT_SCALE_FACTORS_NUM[VPX_SS_MAX_LAYERS] = {
4, 5, 7, 11, 16
};
static const int DEFAULT_SCALE_FACTORS_DEN[VPX_SS_MAX_LAYERS] = {
16, 16, 16, 16, 16
};
typedef enum {
QUANTIZER = 0,
BITRATE,
SCALE_FACTOR,
AUTO_ALT_REF,
ALL_OPTION_TYPES
} LAYER_OPTION_TYPE;
static const int option_max_values[ALL_OPTION_TYPES] = {
63, INT_MAX, INT_MAX, 1
};
static const int option_min_values[ALL_OPTION_TYPES] = {
0, 0, 1, 0
};
// One encoded frame
typedef struct FrameData {
void *buf; // compressed data buffer
size_t size; // length of compressed data
vpx_codec_frame_flags_t flags; /**< flags for this frame */
struct FrameData *next;
} FrameData;
static SvcInternal_t *get_svc_internal(SvcContext *svc_ctx) {
if (svc_ctx == NULL) return NULL;
if (svc_ctx->internal == NULL) {
SvcInternal_t *const si = (SvcInternal_t *)malloc(sizeof(*si));
if (si != NULL) {
memset(si, 0, sizeof(*si));
}
svc_ctx->internal = si;
}
return (SvcInternal_t *)svc_ctx->internal;
}
static const SvcInternal_t *get_const_svc_internal(
const SvcContext *svc_ctx) {
if (svc_ctx == NULL) return NULL;
return (const SvcInternal_t *)svc_ctx->internal;
}
static void svc_log_reset(SvcContext *svc_ctx) {
SvcInternal_t *const si = (SvcInternal_t *)svc_ctx->internal;
si->message_buffer[0] = '\0';
}
static int svc_log(SvcContext *svc_ctx, SVC_LOG_LEVEL level,
const char *fmt, ...) {
char buf[512];
int retval = 0;
va_list ap;
SvcInternal_t *const si = get_svc_internal(svc_ctx);
if (level > svc_ctx->log_level) {
return retval;
}
va_start(ap, fmt);
retval = vsnprintf(buf, sizeof(buf), fmt, ap);
va_end(ap);
if (svc_ctx->log_print) {
printf("%s", buf);
} else {
strncat(si->message_buffer, buf,
sizeof(si->message_buffer) - strlen(si->message_buffer) - 1);
}
if (level == SVC_LOG_ERROR) {
si->codec_ctx->err_detail = si->message_buffer;
}
return retval;
}
static vpx_codec_err_t extract_option(LAYER_OPTION_TYPE type,
char *input,
int *value0,
int *value1) {
if (type == SCALE_FACTOR) {
*value0 = strtol(input, &input, 10);
if (*input++ != '/')
return VPX_CODEC_INVALID_PARAM;
*value1 = strtol(input, &input, 10);
if (*value0 < option_min_values[SCALE_FACTOR] ||
*value1 < option_min_values[SCALE_FACTOR] ||
*value0 > option_max_values[SCALE_FACTOR] ||
*value1 > option_max_values[SCALE_FACTOR] ||
*value0 > *value1) // num shouldn't be greater than den
return VPX_CODEC_INVALID_PARAM;
} else {
*value0 = atoi(input);
if (*value0 < option_min_values[type] ||
*value0 > option_max_values[type])
return VPX_CODEC_INVALID_PARAM;
}
return VPX_CODEC_OK;
}
static vpx_codec_err_t parse_layer_options_from_string(SvcContext *svc_ctx,
LAYER_OPTION_TYPE type,
const char *input,
int *option0,
int *option1) {
int i;
vpx_codec_err_t res = VPX_CODEC_OK;
char *input_string;
char *token;
const char *delim = ",";
char *save_ptr;
if (input == NULL || option0 == NULL ||
(option1 == NULL && type == SCALE_FACTOR))
return VPX_CODEC_INVALID_PARAM;
input_string = strdup(input);
token = strtok_r(input_string, delim, &save_ptr);
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
if (token != NULL) {
res = extract_option(type, token, option0 + i, option1 + i);
if (res != VPX_CODEC_OK)
break;
token = strtok_r(NULL, delim, &save_ptr);
} else {
break;
}
}
if (res == VPX_CODEC_OK && i != svc_ctx->spatial_layers) {
svc_log(svc_ctx, SVC_LOG_ERROR,
"svc: layer params type: %d %d values required, "
"but only %d specified\n", type, svc_ctx->spatial_layers, i);
res = VPX_CODEC_INVALID_PARAM;
}
free(input_string);
return res;
}
/**
* Parse SVC encoding options
* Format: encoding-mode=<svc_mode>,layers=<layer_count>
* scale-factors=<n1>/<d1>,<n2>/<d2>,...
* quantizers=<q1>,<q2>,...
* svc_mode = [i|ip|alt_ip|gf]
*/
static vpx_codec_err_t parse_options(SvcContext *svc_ctx, const char *options) {
char *input_string;
char *option_name;
char *option_value;
char *input_ptr;
SvcInternal_t *const si = get_svc_internal(svc_ctx);
vpx_codec_err_t res = VPX_CODEC_OK;
int i, alt_ref_enabled = 0;
if (options == NULL) return VPX_CODEC_OK;
input_string = strdup(options);
// parse option name
option_name = strtok_r(input_string, "=", &input_ptr);
while (option_name != NULL) {
// parse option value
option_value = strtok_r(NULL, " ", &input_ptr);
if (option_value == NULL) {
svc_log(svc_ctx, SVC_LOG_ERROR, "option missing value: %s\n",
option_name);
res = VPX_CODEC_INVALID_PARAM;
break;
}
if (strcmp("spatial-layers", option_name) == 0) {
svc_ctx->spatial_layers = atoi(option_value);
} else if (strcmp("temporal-layers", option_name) == 0) {
svc_ctx->temporal_layers = atoi(option_value);
} else if (strcmp("scale-factors", option_name) == 0) {
res = parse_layer_options_from_string(svc_ctx, SCALE_FACTOR, option_value,
si->svc_params.scaling_factor_num,
si->svc_params.scaling_factor_den);
if (res != VPX_CODEC_OK) break;
} else if (strcmp("max-quantizers", option_name) == 0) {
res = parse_layer_options_from_string(svc_ctx, QUANTIZER, option_value,
si->svc_params.max_quantizers,
NULL);
if (res != VPX_CODEC_OK) break;
} else if (strcmp("min-quantizers", option_name) == 0) {
res = parse_layer_options_from_string(svc_ctx, QUANTIZER, option_value,
si->svc_params.min_quantizers,
NULL);
if (res != VPX_CODEC_OK) break;
} else if (strcmp("auto-alt-refs", option_name) == 0) {
res = parse_layer_options_from_string(svc_ctx, AUTO_ALT_REF, option_value,
si->enable_auto_alt_ref, NULL);
if (res != VPX_CODEC_OK) break;
} else if (strcmp("bitrates", option_name) == 0) {
res = parse_layer_options_from_string(svc_ctx, BITRATE, option_value,
si->bitrates, NULL);
if (res != VPX_CODEC_OK) break;
} else if (strcmp("multi-frame-contexts", option_name) == 0) {
si->use_multiple_frame_contexts = atoi(option_value);
} else {
svc_log(svc_ctx, SVC_LOG_ERROR, "invalid option: %s\n", option_name);
res = VPX_CODEC_INVALID_PARAM;
break;
}
option_name = strtok_r(NULL, "=", &input_ptr);
}
free(input_string);
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
if (si->svc_params.max_quantizers[i] > MAX_QUANTIZER ||
si->svc_params.max_quantizers[i] < 0 ||
si->svc_params.min_quantizers[i] > si->svc_params.max_quantizers[i] ||
si->svc_params.min_quantizers[i] < 0)
res = VPX_CODEC_INVALID_PARAM;
}
if (si->use_multiple_frame_contexts &&
(svc_ctx->spatial_layers > 3 ||
svc_ctx->spatial_layers * svc_ctx->temporal_layers > 4))
res = VPX_CODEC_INVALID_PARAM;
for (i = 0; i < svc_ctx->spatial_layers; ++i)
alt_ref_enabled += si->enable_auto_alt_ref[i];
if (alt_ref_enabled > REF_FRAMES - svc_ctx->spatial_layers) {
svc_log(svc_ctx, SVC_LOG_ERROR,
"svc: auto alt ref: Maxinum %d(REF_FRAMES - layers) layers could"
"enabled auto alt reference frame, but % layers are enabled\n",
REF_FRAMES - svc_ctx->spatial_layers, alt_ref_enabled);
res = VPX_CODEC_INVALID_PARAM;
}
return res;
}
vpx_codec_err_t vpx_svc_set_options(SvcContext *svc_ctx,
const char *options) {
SvcInternal_t *const si = get_svc_internal(svc_ctx);
if (svc_ctx == NULL || options == NULL || si == NULL) {
return VPX_CODEC_INVALID_PARAM;
}
strncpy(si->options, options, sizeof(si->options));
si->options[sizeof(si->options) - 1] = '\0';
return VPX_CODEC_OK;
}
void assign_layer_bitrates(const SvcContext *svc_ctx,
vpx_codec_enc_cfg_t *const enc_cfg) {
int i;
const SvcInternal_t *const si = get_const_svc_internal(svc_ctx);
int sl, tl, spatial_layer_target;
if (svc_ctx->temporal_layering_mode != 0) {
if (si->bitrates[0] != 0) {
enc_cfg->rc_target_bitrate = 0;
for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
enc_cfg->ss_target_bitrate[sl*svc_ctx->temporal_layers] = 0;
for (tl = 0; tl < svc_ctx->temporal_layers; ++tl) {
enc_cfg->ss_target_bitrate[sl*svc_ctx->temporal_layers]
+= (unsigned int)si->bitrates[sl * svc_ctx->temporal_layers + tl];
enc_cfg->layer_target_bitrate[sl*svc_ctx->temporal_layers + tl]
= si->bitrates[sl * svc_ctx->temporal_layers + tl];
}
}
} else {
float total = 0;
float alloc_ratio[VPX_MAX_LAYERS] = {0};
for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
if (si->svc_params.scaling_factor_den[sl] > 0) {
alloc_ratio[sl] = (float)( pow(2, sl) );
total += alloc_ratio[sl];
}
}
for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
enc_cfg->ss_target_bitrate[sl] = spatial_layer_target =
(unsigned int)(enc_cfg->rc_target_bitrate *
alloc_ratio[sl] / total);
if (svc_ctx->temporal_layering_mode == 3) {
enc_cfg->layer_target_bitrate[sl * svc_ctx->temporal_layers] =
(spatial_layer_target*6)/10; // 60%
enc_cfg->layer_target_bitrate[sl * svc_ctx->temporal_layers + 1] =
(spatial_layer_target*8)/10; // 80%
enc_cfg->layer_target_bitrate[sl * svc_ctx->temporal_layers + 2] =
spatial_layer_target;
} else if (svc_ctx->temporal_layering_mode == 2 ||
svc_ctx->temporal_layering_mode == 1) {
enc_cfg->layer_target_bitrate[sl * svc_ctx->temporal_layers] =
spatial_layer_target * 2 / 3;
enc_cfg->layer_target_bitrate[sl * svc_ctx->temporal_layers + 1] =
spatial_layer_target;
} else {
// User should explicitly assign bitrates in this case.
assert(0);
}
}
}
} else {
if (si->bitrates[0] != 0) {
enc_cfg->rc_target_bitrate = 0;
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
enc_cfg->ss_target_bitrate[i] = (unsigned int)si->bitrates[i];
enc_cfg->rc_target_bitrate += si->bitrates[i];
}
} else {
float total = 0;
float alloc_ratio[VPX_MAX_LAYERS] = {0};
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
if (si->svc_params.scaling_factor_den[i] > 0) {
alloc_ratio[i] = (float)(si->svc_params.scaling_factor_num[i] * 1.0 /
si->svc_params.scaling_factor_den[i]);
alloc_ratio[i] *= alloc_ratio[i];
total += alloc_ratio[i];
}
}
for (i = 0; i < VPX_SS_MAX_LAYERS; ++i) {
if (total > 0) {
enc_cfg->layer_target_bitrate[i] = (unsigned int)
(enc_cfg->rc_target_bitrate * alloc_ratio[i] / total);
}
}
}
}
}
vpx_codec_err_t vpx_svc_init(SvcContext *svc_ctx, vpx_codec_ctx_t *codec_ctx,
vpx_codec_iface_t *iface,
vpx_codec_enc_cfg_t *enc_cfg) {
vpx_codec_err_t res;
int i, sl , tl;
SvcInternal_t *const si = get_svc_internal(svc_ctx);
if (svc_ctx == NULL || codec_ctx == NULL || iface == NULL ||
enc_cfg == NULL) {
return VPX_CODEC_INVALID_PARAM;
}
if (si == NULL) return VPX_CODEC_MEM_ERROR;
si->codec_ctx = codec_ctx;
si->width = enc_cfg->g_w;
si->height = enc_cfg->g_h;
si->kf_dist = enc_cfg->kf_max_dist;
if (svc_ctx->spatial_layers == 0)
svc_ctx->spatial_layers = VPX_SS_DEFAULT_LAYERS;
if (svc_ctx->spatial_layers < 1 ||
svc_ctx->spatial_layers > VPX_SS_MAX_LAYERS) {
svc_log(svc_ctx, SVC_LOG_ERROR, "spatial layers: invalid value: %d\n",
svc_ctx->spatial_layers);
return VPX_CODEC_INVALID_PARAM;
}
// Note: temporal_layering_mode only applies to one-pass CBR
// si->svc_params.temporal_layering_mode = svc_ctx->temporal_layering_mode;
if (svc_ctx->temporal_layering_mode == 3) {
svc_ctx->temporal_layers = 3;
} else if (svc_ctx->temporal_layering_mode == 2 ||
svc_ctx->temporal_layering_mode == 1) {
svc_ctx->temporal_layers = 2;
}
for (sl = 0; sl < VPX_SS_MAX_LAYERS; ++sl) {
si->svc_params.scaling_factor_num[sl] = DEFAULT_SCALE_FACTORS_NUM[sl];
si->svc_params.scaling_factor_den[sl] = DEFAULT_SCALE_FACTORS_DEN[sl];
}
for (tl = 0; tl < svc_ctx->temporal_layers; ++tl) {
for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
i = sl * svc_ctx->temporal_layers + tl;
si->svc_params.max_quantizers[i] = MAX_QUANTIZER;
si->svc_params.min_quantizers[i] = 0;
}
}
// Parse aggregate command line options. Options must start with
// "layers=xx" then followed by other options
res = parse_options(svc_ctx, si->options);
if (res != VPX_CODEC_OK) return res;
if (svc_ctx->spatial_layers < 1)
svc_ctx->spatial_layers = 1;
if (svc_ctx->spatial_layers > VPX_SS_MAX_LAYERS)
svc_ctx->spatial_layers = VPX_SS_MAX_LAYERS;
if (svc_ctx->temporal_layers < 1)
svc_ctx->temporal_layers = 1;
if (svc_ctx->temporal_layers > VPX_TS_MAX_LAYERS)
svc_ctx->temporal_layers = VPX_TS_MAX_LAYERS;
if (svc_ctx->temporal_layers * svc_ctx->spatial_layers > VPX_MAX_LAYERS) {
svc_log(svc_ctx, SVC_LOG_ERROR,
"spatial layers * temporal layers exceeds the maximum number of "
"allowed layers of %d\n",
svc_ctx->spatial_layers * svc_ctx->temporal_layers,
(int) VPX_MAX_LAYERS);
return VPX_CODEC_INVALID_PARAM;
}
assign_layer_bitrates(svc_ctx, enc_cfg);
#if CONFIG_SPATIAL_SVC
for (i = 0; i < svc_ctx->spatial_layers; ++i)
enc_cfg->ss_enable_auto_alt_ref[i] = si->enable_auto_alt_ref[i];
#endif
if (svc_ctx->temporal_layers > 1) {
int i;
for (i = 0; i < svc_ctx->temporal_layers; ++i) {
enc_cfg->ts_target_bitrate[i] = enc_cfg->rc_target_bitrate /
svc_ctx->temporal_layers;
enc_cfg->ts_rate_decimator[i] = 1 << (svc_ctx->temporal_layers - 1 - i);
}
}
if (svc_ctx->threads)
enc_cfg->g_threads = svc_ctx->threads;
// Modify encoder configuration
enc_cfg->ss_number_layers = svc_ctx->spatial_layers;
enc_cfg->ts_number_layers = svc_ctx->temporal_layers;
if (enc_cfg->rc_end_usage == VPX_CBR) {
enc_cfg->rc_resize_allowed = 0;
enc_cfg->rc_min_quantizer = 2;
enc_cfg->rc_max_quantizer = 56;
enc_cfg->rc_undershoot_pct = 50;
enc_cfg->rc_overshoot_pct = 50;
enc_cfg->rc_buf_initial_sz = 500;
enc_cfg->rc_buf_optimal_sz = 600;
enc_cfg->rc_buf_sz = 1000;
enc_cfg->rc_dropframe_thresh = 0;
}
if (enc_cfg->g_error_resilient == 0 && si->use_multiple_frame_contexts == 0)
enc_cfg->g_error_resilient = 1;
// Initialize codec
res = vpx_codec_enc_init(codec_ctx, iface, enc_cfg, VPX_CODEC_USE_PSNR);
if (res != VPX_CODEC_OK) {
svc_log(svc_ctx, SVC_LOG_ERROR, "svc_enc_init error\n");
return res;
}
if (svc_ctx->spatial_layers > 1 || svc_ctx->temporal_layers > 1) {
vpx_codec_control(codec_ctx, VP9E_SET_SVC, 1);
vpx_codec_control(codec_ctx, VP9E_SET_SVC_PARAMETERS, &si->svc_params);
}
return VPX_CODEC_OK;
}
/**
* Encode a frame into multiple layers
* Create a superframe containing the individual layers
*/
vpx_codec_err_t vpx_svc_encode(SvcContext *svc_ctx,
vpx_codec_ctx_t *codec_ctx,
struct vpx_image *rawimg,
vpx_codec_pts_t pts,
int64_t duration, int deadline) {
vpx_codec_err_t res;
vpx_codec_iter_t iter;
const vpx_codec_cx_pkt_t *cx_pkt;
SvcInternal_t *const si = get_svc_internal(svc_ctx);
if (svc_ctx == NULL || codec_ctx == NULL || si == NULL) {
return VPX_CODEC_INVALID_PARAM;
}
svc_log_reset(svc_ctx);
res = vpx_codec_encode(codec_ctx, rawimg, pts, (uint32_t)duration, 0,
deadline);
if (res != VPX_CODEC_OK) {
return res;
}
// save compressed data
iter = NULL;
while ((cx_pkt = vpx_codec_get_cx_data(codec_ctx, &iter))) {
switch (cx_pkt->kind) {
#if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
#if CONFIG_SPATIAL_SVC
case VPX_CODEC_SPATIAL_SVC_LAYER_PSNR: {
int i;
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
int j;
svc_log(svc_ctx, SVC_LOG_DEBUG,
"SVC frame: %d, layer: %d, PSNR(Total/Y/U/V): "
"%2.3f %2.3f %2.3f %2.3f \n",
si->psnr_pkt_received, i,
cx_pkt->data.layer_psnr[i].psnr[0],
cx_pkt->data.layer_psnr[i].psnr[1],
cx_pkt->data.layer_psnr[i].psnr[2],
cx_pkt->data.layer_psnr[i].psnr[3]);
svc_log(svc_ctx, SVC_LOG_DEBUG,
"SVC frame: %d, layer: %d, SSE(Total/Y/U/V): "
"%2.3f %2.3f %2.3f %2.3f \n",
si->psnr_pkt_received, i,
cx_pkt->data.layer_psnr[i].sse[0],
cx_pkt->data.layer_psnr[i].sse[1],
cx_pkt->data.layer_psnr[i].sse[2],
cx_pkt->data.layer_psnr[i].sse[3]);
for (j = 0; j < COMPONENTS; ++j) {
si->psnr_sum[i][j] +=
cx_pkt->data.layer_psnr[i].psnr[j];
si->sse_sum[i][j] += cx_pkt->data.layer_psnr[i].sse[j];
}
}
++si->psnr_pkt_received;
break;
}
case VPX_CODEC_SPATIAL_SVC_LAYER_SIZES: {
int i;
for (i = 0; i < svc_ctx->spatial_layers; ++i)
si->bytes_sum[i] += cx_pkt->data.layer_sizes[i];
break;
}
#endif
#endif
case VPX_CODEC_PSNR_PKT:
{
#if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
int j;
svc_log(svc_ctx, SVC_LOG_DEBUG,
"frame: %d, layer: %d, PSNR(Total/Y/U/V): "
"%2.3f %2.3f %2.3f %2.3f \n",
si->psnr_pkt_received, 0,
cx_pkt->data.layer_psnr[0].psnr[0],
cx_pkt->data.layer_psnr[0].psnr[1],
cx_pkt->data.layer_psnr[0].psnr[2],
cx_pkt->data.layer_psnr[0].psnr[3]);
for (j = 0; j < COMPONENTS; ++j) {
si->psnr_sum[0][j] +=
cx_pkt->data.layer_psnr[0].psnr[j];
si->sse_sum[0][j] += cx_pkt->data.layer_psnr[0].sse[j];
}
#endif
}
++si->psnr_pkt_received;
break;
default: {
break;
}
}
}
return VPX_CODEC_OK;
}
const char *vpx_svc_get_message(const SvcContext *svc_ctx) {
const SvcInternal_t *const si = get_const_svc_internal(svc_ctx);
if (svc_ctx == NULL || si == NULL) return NULL;
return si->message_buffer;
}
static double calc_psnr(double d) {
if (d == 0) return 100;
return -10.0 * log(d) / log(10.0);
}
// dump accumulated statistics and reset accumulated values
const char *vpx_svc_dump_statistics(SvcContext *svc_ctx) {
int number_of_frames;
int i, j;
uint32_t bytes_total = 0;
double scale[COMPONENTS];
double psnr[COMPONENTS];
double mse[COMPONENTS];
double y_scale;
SvcInternal_t *const si = get_svc_internal(svc_ctx);
if (svc_ctx == NULL || si == NULL) return NULL;
svc_log_reset(svc_ctx);
number_of_frames = si->psnr_pkt_received;
if (number_of_frames <= 0) return vpx_svc_get_message(svc_ctx);
svc_log(svc_ctx, SVC_LOG_INFO, "\n");
for (i = 0; i < svc_ctx->spatial_layers; ++i) {
svc_log(svc_ctx, SVC_LOG_INFO,
"Layer %d Average PSNR=[%2.3f, %2.3f, %2.3f, %2.3f], Bytes=[%u]\n",
i, (double)si->psnr_sum[i][0] / number_of_frames,
(double)si->psnr_sum[i][1] / number_of_frames,
(double)si->psnr_sum[i][2] / number_of_frames,
(double)si->psnr_sum[i][3] / number_of_frames, si->bytes_sum[i]);
// the following psnr calculation is deduced from ffmpeg.c#print_report
y_scale = si->width * si->height * 255.0 * 255.0 * number_of_frames;
scale[1] = y_scale;
scale[2] = scale[3] = y_scale / 4; // U or V
scale[0] = y_scale * 1.5; // total
for (j = 0; j < COMPONENTS; j++) {
psnr[j] = calc_psnr(si->sse_sum[i][j] / scale[j]);
mse[j] = si->sse_sum[i][j] * 255.0 * 255.0 / scale[j];
}
svc_log(svc_ctx, SVC_LOG_INFO,
"Layer %d Overall PSNR=[%2.3f, %2.3f, %2.3f, %2.3f]\n", i, psnr[0],
psnr[1], psnr[2], psnr[3]);
svc_log(svc_ctx, SVC_LOG_INFO,
"Layer %d Overall MSE=[%2.3f, %2.3f, %2.3f, %2.3f]\n", i, mse[0],
mse[1], mse[2], mse[3]);
bytes_total += si->bytes_sum[i];
// Clear sums for next time.
si->bytes_sum[i] = 0;
for (j = 0; j < COMPONENTS; ++j) {
si->psnr_sum[i][j] = 0;
si->sse_sum[i][j] = 0;
}
}
// only display statistics once
si->psnr_pkt_received = 0;
svc_log(svc_ctx, SVC_LOG_INFO, "Total Bytes=[%u]\n", bytes_total);
return vpx_svc_get_message(svc_ctx);
}
void vpx_svc_release(SvcContext *svc_ctx) {
SvcInternal_t *si;
if (svc_ctx == NULL) return;
// do not use get_svc_internal as it will unnecessarily allocate an
// SvcInternal_t if it was not already allocated
si = (SvcInternal_t *)svc_ctx->internal;
if (si != NULL) {
free(si);
svc_ctx->internal = NULL;
}
}

123
vpx/svc_context.h
Просмотреть файл

@ -1,123 +0,0 @@
/*
* Copyright (c) 2013 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
/**
* SvcContext - input parameters and state to encode a multi-layered
* spatial SVC frame
*/
#ifndef VPX_SVC_CONTEXT_H_
#define VPX_SVC_CONTEXT_H_
#include "./vp8cx.h"
#include "./vpx_encoder.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef enum SVC_LOG_LEVEL {
SVC_LOG_ERROR,
SVC_LOG_INFO,
SVC_LOG_DEBUG
} SVC_LOG_LEVEL;
typedef struct {
// public interface to svc_command options
int spatial_layers; // number of spatial layers
int temporal_layers; // number of temporal layers
int temporal_layering_mode;
SVC_LOG_LEVEL log_level; // amount of information to display
int log_print; // when set, printf log messages instead of returning the
// message with svc_get_message
int output_rc_stat; // for outputting rc stats
int speed; // speed setting for codec
int threads;
int aqmode; // turns on aq-mode=3 (cyclic_refresh): 0=off, 1=on.
// private storage for vpx_svc_encode
void *internal;
} SvcContext;
#define OPTION_BUFFER_SIZE 1024
#define COMPONENTS 4 // psnr & sse statistics maintained for total, y, u, v
typedef struct SvcInternal {
char options[OPTION_BUFFER_SIZE]; // set by vpx_svc_set_options
// values extracted from option, quantizers
vpx_svc_extra_cfg_t svc_params;
int enable_auto_alt_ref[VPX_SS_MAX_LAYERS];
int bitrates[VPX_SS_MAX_LAYERS];
// accumulated statistics
double psnr_sum[VPX_SS_MAX_LAYERS][COMPONENTS]; // total/Y/U/V
uint64_t sse_sum[VPX_SS_MAX_LAYERS][COMPONENTS];
uint32_t bytes_sum[VPX_SS_MAX_LAYERS];
// codec encoding values
int width; // width of highest layer
int height; // height of highest layer
int kf_dist; // distance between keyframes
// state variables
int psnr_pkt_received;
int layer;
int use_multiple_frame_contexts;
char message_buffer[2048];
vpx_codec_ctx_t *codec_ctx;
} SvcInternal_t;
/**
* Set SVC options
* options are supplied as a single string separated by spaces
* Format: encoding-mode=<i|ip|alt-ip|gf>
* layers=<layer_count>
* scaling-factors=<n1>/<d1>,<n2>/<d2>,...
* quantizers=<q1>,<q2>,...
*/
vpx_codec_err_t vpx_svc_set_options(SvcContext *svc_ctx, const char *options);
/**
* initialize SVC encoding
*/
vpx_codec_err_t vpx_svc_init(SvcContext *svc_ctx,
vpx_codec_ctx_t *codec_ctx,
vpx_codec_iface_t *iface,
vpx_codec_enc_cfg_t *cfg);
/**
* encode a frame of video with multiple layers
*/
vpx_codec_err_t vpx_svc_encode(SvcContext *svc_ctx,
vpx_codec_ctx_t *codec_ctx,
struct vpx_image *rawimg,
vpx_codec_pts_t pts,
int64_t duration, int deadline);
/**
* finished with svc encoding, release allocated resources
*/
void vpx_svc_release(SvcContext *svc_ctx);
/**
* dump accumulated statistics and reset accumulated values
*/
const char *vpx_svc_dump_statistics(SvcContext *svc_ctx);
/**
* get status message from previous encode
*/
const char *vpx_svc_get_message(const SvcContext *svc_ctx);
#ifdef __cplusplus
} // extern "C"
#endif
#endif // VPX_SVC_CONTEXT_H_

142
vpx/vp8cx.h
Просмотреть файл

@ -301,17 +301,6 @@ enum vp8e_enc_control_id {
*/
VP9E_SET_GF_CBR_BOOST_PCT,
/*!\brief Codec control function to set the temporal layer id.
*
* For temporal scalability: this control allows the application to set the
* layer id for each frame to be encoded. Note that this control must be set
* for every frame prior to encoding. The usage of this control function
* supersedes the internal temporal pattern counter, which is now deprecated.
*
* Supported in codecs: VP8
*/
VP8E_SET_TEMPORAL_LAYER_ID,
/*!\brief Codec control function to set encoder screen content mode.
*
* 0: off, 1: On, 2: On with more aggressive rate control.
@ -428,32 +417,6 @@ enum vp8e_enc_control_id {
*/
VP9E_SET_NOISE_SENSITIVITY,
/*!\brief Codec control function to turn on/off SVC in encoder.
* \note Return value is VPX_CODEC_INVALID_PARAM if the encoder does not
* support SVC in its current encoding mode
* 0: off, 1: on
*
* Supported in codecs: VP9
*/
VP9E_SET_SVC,
/*!\brief Codec control function to set parameters for SVC.
* \note Parameters contain min_q, max_q, scaling factor for each of the
* SVC layers.
*
* Supported in codecs: VP9
*/
VP9E_SET_SVC_PARAMETERS,
/*!\brief Codec control function to set svc layer for spatial and temporal.
* \note Valid ranges: 0..#vpx_codec_enc_cfg::ss_number_layers for spatial
* layer and 0..#vpx_codec_enc_cfg::ts_number_layers for
* temporal layer.
*
* Supported in codecs: VP9
*/
VP9E_SET_SVC_LAYER_ID,
/*!\brief Codec control function to set content type.
* \note Valid parameter range:
* VPX_CONTENT_DEFAULT = Regular video content (Default)
@ -463,22 +426,6 @@ enum vp8e_enc_control_id {
*/
VP9E_SET_TUNE_CONTENT,
/*!\brief Codec control function to get svc layer ID.
* \note The layer ID returned is for the data packet from the registered
* callback function.
*
* Supported in codecs: VP9
*/
VP9E_GET_SVC_LAYER_ID,
/*!\brief Codec control function to register callback to get per layer packet.
* \note Parameter for this control function is a structure with a callback
* function and a pointer to private data used by the callback.
*
* Supported in codecs: VP9
*/
VP9E_REGISTER_CX_CALLBACK,
/*!\brief Codec control function to set color space info.
* \note Valid ranges: 0..7, default is "UNKNOWN".
* 0 = UNKNOWN,
@ -494,17 +441,6 @@ enum vp8e_enc_control_id {
*/
VP9E_SET_COLOR_SPACE,
/*!\brief Codec control function to set temporal layering mode.
* \note Valid ranges: 0..3, default is "0" (VP9E_TEMPORAL_LAYERING_MODE_NOLAYERING).
* 0 = VP9E_TEMPORAL_LAYERING_MODE_NOLAYERING
* 1 = VP9E_TEMPORAL_LAYERING_MODE_BYPASS
* 2 = VP9E_TEMPORAL_LAYERING_MODE_0101
* 3 = VP9E_TEMPORAL_LAYERING_MODE_0212
*
* Supported in codecs: VP9
*/
VP9E_SET_TEMPORAL_LAYERING_MODE,
/*!\brief Codec control function to set minimum interval between GF/ARF frames
*
* By default the value is set as 4.
@ -536,14 +472,6 @@ enum vp8e_enc_control_id {
*/
VP9E_SET_COLOR_RANGE,
/*!\brief Codec control function to set the frame flags and buffer indices
* for spatial layers. The frame flags and buffer indices are set using the
* struct #vpx_svc_ref_frame_config defined below.
*
* Supported in codecs: VP9
*/
VP9E_SET_SVC_REF_FRAME_CONFIG,
/*!\brief Codec control function to set intended rendering image size.
*
* By default, this is identical to the image size in pixels.
@ -588,33 +516,6 @@ typedef enum vpx_scaling_mode_1d {
VP8E_ONETWO = 3
} VPX_SCALING_MODE;
/*!\brief Temporal layering mode enum for VP9 SVC.
*
* This set of macros define the different temporal layering modes.
* Supported codecs: VP9 (in SVC mode)
*
*/
typedef enum vp9e_temporal_layering_mode {
/*!\brief No temporal layering.
* Used when only spatial layering is used.
*/
VP9E_TEMPORAL_LAYERING_MODE_NOLAYERING = 0,
/*!\brief Bypass mode.
* Used when application needs to control temporal layering.
* This will only work when the number of spatial layers equals 1.
*/
VP9E_TEMPORAL_LAYERING_MODE_BYPASS = 1,
/*!\brief 0-1-0-1... temporal layering scheme with two temporal layers.
*/
VP9E_TEMPORAL_LAYERING_MODE_0101 = 2,
/*!\brief 0-2-1-2... temporal layering scheme with three temporal layers.
*/
VP9E_TEMPORAL_LAYERING_MODE_0212 = 3
} VP9E_TEMPORAL_LAYERING_MODE;
/*!\brief vpx region of interest map
*
* These defines the data structures for the region of interest map
@ -689,33 +590,6 @@ typedef enum {
VPX_TUNE_SSIM
} vpx_tune_metric;
/*!\brief vp9 svc layer parameters
*
* This defines the spatial and temporal layer id numbers for svc encoding.
* This is used with the #VP9E_SET_SVC_LAYER_ID control to set the spatial and
* temporal layer id for the current frame.
*
*/
typedef struct vpx_svc_layer_id {
int spatial_layer_id; /**< Spatial layer id number. */
int temporal_layer_id; /**< Temporal layer id number. */
} vpx_svc_layer_id_t;
/*!\brief vp9 svc frame flag parameters.
*
* This defines the frame flags and buffer indices for each spatial layer for
* svc encoding.
* This is used with the #VP9E_SET_SVC_REF_FRAME_CONFIG control to set frame
* flags and buffer indices for each spatial layer for the current (super)frame.
*
*/
typedef struct vpx_svc_ref_frame_config {
int frame_flags[VPX_TS_MAX_LAYERS]; /**< Frame flags. */
int lst_fb_idx[VPX_TS_MAX_LAYERS]; /**< Last buffer index. */
int gld_fb_idx[VPX_TS_MAX_LAYERS]; /**< Golden buffer index. */
int alt_fb_idx[VPX_TS_MAX_LAYERS]; /**< Altref buffer index. */
} vpx_svc_ref_frame_config_t;
/*!\cond */
/*!\brief VP8 encoder control function parameter type
*
@ -728,8 +602,6 @@ VPX_CTRL_USE_TYPE_DEPRECATED(VP8E_USE_REFERENCE, int)
#define VPX_CTRL_VP8E_USE_REFERENCE
VPX_CTRL_USE_TYPE(VP8E_SET_FRAME_FLAGS, int)
#define VPX_CTRL_VP8E_SET_FRAME_FLAGS
VPX_CTRL_USE_TYPE(VP8E_SET_TEMPORAL_LAYER_ID, int)
#define VPX_CTRL_VP8E_SET_TEMPORAL_LAYER_ID
VPX_CTRL_USE_TYPE(VP8E_SET_ROI_MAP, vpx_roi_map_t *)
#define VPX_CTRL_VP8E_SET_ROI_MAP
VPX_CTRL_USE_TYPE(VP8E_SET_ACTIVEMAP, vpx_active_map_t *)
@ -737,15 +609,6 @@ VPX_CTRL_USE_TYPE(VP8E_SET_ACTIVEMAP, vpx_active_map_t *)
VPX_CTRL_USE_TYPE(VP8E_SET_SCALEMODE, vpx_scaling_mode_t *)
#define VPX_CTRL_VP8E_SET_SCALEMODE
VPX_CTRL_USE_TYPE(VP9E_SET_SVC, int)
#define VPX_CTRL_VP9E_SET_SVC
VPX_CTRL_USE_TYPE(VP9E_SET_SVC_PARAMETERS, void *)
#define VPX_CTRL_VP9E_SET_SVC_PARAMETERS
VPX_CTRL_USE_TYPE(VP9E_REGISTER_CX_CALLBACK, void *)
#define VPX_CTRL_VP9E_REGISTER_CX_CALLBACK
VPX_CTRL_USE_TYPE(VP9E_SET_SVC_LAYER_ID, vpx_svc_layer_id_t *)
#define VPX_CTRL_VP9E_SET_SVC_LAYER_ID
VPX_CTRL_USE_TYPE(VP8E_SET_CPUUSED, int)
#define VPX_CTRL_VP8E_SET_CPUUSED
VPX_CTRL_USE_TYPE(VP8E_SET_ENABLEAUTOALTREF, unsigned int)
@ -785,8 +648,6 @@ VPX_CTRL_USE_TYPE(VP8E_GET_LAST_QUANTIZER, int *)
#define VPX_CTRL_VP8E_GET_LAST_QUANTIZER
VPX_CTRL_USE_TYPE(VP8E_GET_LAST_QUANTIZER_64, int *)
#define VPX_CTRL_VP8E_GET_LAST_QUANTIZER_64
VPX_CTRL_USE_TYPE(VP9E_GET_SVC_LAYER_ID, vpx_svc_layer_id_t *)
#define VPX_CTRL_VP9E_GET_SVC_LAYER_ID
VPX_CTRL_USE_TYPE(VP8E_SET_MAX_INTRA_BITRATE_PCT, unsigned int)
#define VPX_CTRL_VP8E_SET_MAX_INTRA_BITRATE_PCT
@ -832,9 +693,6 @@ VPX_CTRL_USE_TYPE(VP9E_GET_ACTIVEMAP, vpx_active_map_t *)
VPX_CTRL_USE_TYPE(VP9E_SET_COLOR_RANGE, int)
#define VPX_CTRL_VP9E_SET_COLOR_RANGE
VPX_CTRL_USE_TYPE(VP9E_SET_SVC_REF_FRAME_CONFIG, vpx_svc_ref_frame_config_t *)
#define VPX_CTRL_VP9E_SET_SVC_REF_FRAME_CONFIG
/*!\brief
*
* TODO(rbultje) : add support of the control in ffmpeg

144
vpx/vpx_encoder.h
Просмотреть файл

@ -31,29 +31,6 @@ extern "C" {
#include "./vpx_codec.h"
/*! Temporal Scalability: Maximum length of the sequence defining frame
* layer membership
*/
#define VPX_TS_MAX_PERIODICITY 16
/*! Temporal Scalability: Maximum number of coding layers */
#define VPX_TS_MAX_LAYERS 5
/*!\deprecated Use #VPX_TS_MAX_PERIODICITY instead. */
#define MAX_PERIODICITY VPX_TS_MAX_PERIODICITY
/*! Temporal+Spatial Scalability: Maximum number of coding layers */
#define VPX_MAX_LAYERS 12 // 3 temporal + 4 spatial layers are allowed.
/*!\deprecated Use #VPX_MAX_LAYERS instead. */
#define MAX_LAYERS VPX_MAX_LAYERS // 3 temporal + 4 spatial layers allowed.
/*! Spatial Scalability: Maximum number of coding layers */
#define VPX_SS_MAX_LAYERS 5
/*! Spatial Scalability: Default number of coding layers */
#define VPX_SS_DEFAULT_LAYERS 1
/*!\brief Current ABI version number
*
* \internal
@ -164,12 +141,6 @@ extern "C" {
VPX_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */
VPX_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */
VPX_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */
// Spatial SVC is still experimental and may be removed before the next ABI
// bump.
#if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
VPX_CODEC_SPATIAL_SVC_LAYER_SIZES, /**< Sizes for each layer in this frame*/
VPX_CODEC_SPATIAL_SVC_LAYER_PSNR, /**< PSNR for each layer in this frame*/
#endif
VPX_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions */
};
@ -206,12 +177,6 @@ extern "C" {
double psnr[4]; /**< PSNR, total/y/u/v */
} psnr; /**< data for PSNR packet */
vpx_fixed_buf_t raw; /**< data for arbitrary packets */
// Spatial SVC is still experimental and may be removed before the next
// ABI bump.
#if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
size_t layer_sizes[VPX_SS_MAX_LAYERS];
struct vpx_psnr_pkt layer_psnr[VPX_SS_MAX_LAYERS];
#endif
/* This packet size is fixed to allow codecs to extend this
* interface without having to manage storage for raw packets,
@ -222,23 +187,6 @@ extern "C" {
} data; /**< packet data */
} vpx_codec_cx_pkt_t; /**< alias for struct vpx_codec_cx_pkt */
/*!\brief Encoder return output buffer callback
*
* This callback function, when registered, returns with packets when each
* spatial layer is encoded.
*/
// putting the definitions here for now. (agrange: find if there
// is a better place for this)
typedef void (* vpx_codec_enc_output_cx_pkt_cb_fn_t)(vpx_codec_cx_pkt_t *pkt,
void *user_data);
/*!\brief Callback function pointer / user data pair storage */
typedef struct vpx_codec_enc_output_cx_cb_pair {
vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt; /**< Callback function */
void *user_priv; /**< Pointer to private data */
} vpx_codec_priv_output_cx_pkt_cb_pair_t;
/*!\brief Rational Number
*
* This structure holds a fractional value.
@ -670,100 +618,8 @@ extern "C" {
* equal to kf_max_dist for a fixed interval.
*/
unsigned int kf_max_dist;
/*
* Spatial scalability settings (ss)
*/
/*!\brief Number of spatial coding layers.
*
* This value specifies the number of spatial coding layers to be used.
*/
unsigned int ss_number_layers;
/*!\brief Enable auto alt reference flags for each spatial layer.
*
* These values specify if auto alt reference frame is enabled for each
* spatial layer.
*/
int ss_enable_auto_alt_ref[VPX_SS_MAX_LAYERS];
/*!\brief Target bitrate for each spatial layer.
*
* These values specify the target coding bitrate to be used for each
* spatial layer.
*/
unsigned int ss_target_bitrate[VPX_SS_MAX_LAYERS];
/*!\brief Number of temporal coding layers.
*
* This value specifies the number of temporal layers to be used.
*/
unsigned int ts_number_layers;
/*!\brief Target bitrate for each temporal layer.
*
* These values specify the target coding bitrate to be used for each
* temporal layer.
*/
unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS];
/*!\brief Frame rate decimation factor for each temporal layer.
*
* These values specify the frame rate decimation factors to apply
* to each temporal layer.
*/
unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS];
/*!\brief Length of the sequence defining frame temporal layer membership.
*
* This value specifies the length of the sequence that defines the
* membership of frames to temporal layers. For example, if the
* ts_periodicity = 8, then the frames are assigned to coding layers with a
* repeated sequence of length 8.
*/
unsigned int ts_periodicity;
/*!\brief Template defining the membership of frames to temporal layers.
*
* This array defines the membership of frames to temporal coding layers.
* For a 2-layer encoding that assigns even numbered frames to one temporal
* layer (0) and odd numbered frames to a second temporal layer (1) with
* ts_periodicity=8, then ts_layer_id = (0,1,0,1,0,1,0,1).
*/
unsigned int ts_layer_id[VPX_TS_MAX_PERIODICITY];
/*!\brief Target bitrate for each spatial/temporal layer.
*
* These values specify the target coding bitrate to be used for each
* spatial/temporal layer.
*
*/
unsigned int layer_target_bitrate[VPX_MAX_LAYERS];
/*!\brief Temporal layering mode indicating which temporal layering scheme to use.
*
* The value (refer to VP9E_TEMPORAL_LAYERING_MODE) specifies the
* temporal layering mode to use.
*
*/
int temporal_layering_mode;
} vpx_codec_enc_cfg_t; /**< alias for struct vpx_codec_enc_cfg */
/*!\brief vp9 svc extra configure parameters
*
* This defines max/min quantizers and scale factors for each layer
*
*/
typedef struct vpx_svc_parameters {
int max_quantizers[VPX_MAX_LAYERS]; /**< Max Q for each layer */
int min_quantizers[VPX_MAX_LAYERS]; /**< Min Q for each layer */
int scaling_factor_num[VPX_MAX_LAYERS]; /**< Scaling factor-numerator */
int scaling_factor_den[VPX_MAX_LAYERS]; /**< Scaling factor-denominator */
int temporal_layering_mode; /**< Temporal layering mode */
} vpx_svc_extra_cfg_t;
/*!\brief Initialize an encoder instance
*
* Initializes a encoder context using the given interface. Applications