aom/vp9/decoder/vp9_detokenize.c

213 строки
6.8 KiB
C
Исходник Обычный вид История

2010-05-18 19:58:33 +04:00
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
2010-05-18 19:58:33 +04:00
*
* 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.
2010-05-18 19:58:33 +04:00
*/
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
2010-05-18 19:58:33 +04:00
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_common.h"
2011-10-05 14:26:00 +04:00
#include "vp9/decoder/vp9_detokenize.h"
2010-05-18 19:58:33 +04:00
#define EOB_CONTEXT_NODE 0
#define ZERO_CONTEXT_NODE 1
#define ONE_CONTEXT_NODE 2
#define LOW_VAL_CONTEXT_NODE 0
#define TWO_CONTEXT_NODE 1
#define THREE_CONTEXT_NODE 2
#define HIGH_LOW_CONTEXT_NODE 3
#define CAT_ONE_CONTEXT_NODE 4
#define CAT_THREEFOUR_CONTEXT_NODE 5
#define CAT_THREE_CONTEXT_NODE 6
#define CAT_FIVE_CONTEXT_NODE 7
2010-05-18 19:58:33 +04:00
#define CAT1_MIN_VAL 5
#define CAT2_MIN_VAL 7
#define CAT3_MIN_VAL 11
#define CAT4_MIN_VAL 19
#define CAT5_MIN_VAL 35
#define CAT6_MIN_VAL 67
#define CAT1_PROB0 159
#define CAT2_PROB0 145
#define CAT2_PROB1 165
#define CAT3_PROB0 140
#define CAT3_PROB1 148
#define CAT3_PROB2 173
#define CAT4_PROB0 135
#define CAT4_PROB1 140
#define CAT4_PROB2 155
#define CAT4_PROB3 176
#define CAT5_PROB0 130
#define CAT5_PROB1 134
#define CAT5_PROB2 141
#define CAT5_PROB3 157
#define CAT5_PROB4 180
static const vp9_prob cat6_prob[15] = {
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-08 02:45:05 +04:00
254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0
};
2010-05-18 19:58:33 +04:00
#define INCREMENT_COUNT(token) \
do { \
if (!cm->frame_parallel_decoding_mode) \
++coef_counts[band][ctx][token]; \
} while (0)
#define WRITE_COEF_CONTINUE(val, token) \
{ \
v = (val * dqv) >> dq_shift; \
dqcoeff[scan[c]] = vp9_read_bit(r) ? -v : v; \
token_cache[scan[c]] = vp9_pt_energy_class[token]; \
++c; \
ctx = get_coef_context(nb, token_cache, c); \
dqv = dq[1]; \
continue; \
}
2010-05-18 19:58:33 +04:00
#define ADJUST_COEF(prob, bits_count) \
do { \
val += (vp9_read(r, prob) << bits_count); \
} while (0)
2010-05-18 19:58:33 +04:00
static int decode_coefs(VP9_COMMON *cm, const MACROBLOCKD *xd, int block,
PLANE_TYPE type, int16_t *dqcoeff, TX_SIZE tx_size,
const int16_t *dq, int ctx, vp9_reader *r) {
const int max_eob = 16 << (tx_size << 1);
const FRAME_CONTEXT *const fc = &cm->fc;
FRAME_COUNTS *const counts = &cm->counts;
const int ref = is_inter_block(&xd->mi_8x8[0]->mbmi);
int band, c = 0;
const vp9_prob (*coef_probs)[COEFF_CONTEXTS][UNCONSTRAINED_NODES] =
fc->coef_probs[tx_size][type][ref];
const vp9_prob *prob;
unsigned int (*coef_counts)[COEFF_CONTEXTS][UNCONSTRAINED_NODES + 1] =
counts->coef[tx_size][type][ref];
unsigned int (*eob_branch_count)[COEFF_CONTEXTS] =
counts->eob_branch[tx_size][type][ref];
uint8_t token_cache[32 * 32];
const uint8_t *cat6;
const uint8_t *band_translate = get_band_translate(tx_size);
const int dq_shift = (tx_size == TX_32X32);
const scan_order *so = get_scan(xd, tx_size, type, block);
const int16_t *scan = so->scan;
const int16_t *nb = so->neighbors;
int v;
int16_t dqv = dq[0];
while (c < max_eob) {
int val;
band = *band_translate++;
prob = coef_probs[band][ctx];
if (!cm->frame_parallel_decoding_mode)
++eob_branch_count[band][ctx];
if (!vp9_read(r, prob[EOB_CONTEXT_NODE])) {
INCREMENT_COUNT(EOB_MODEL_TOKEN);
break;
}
while (!vp9_read(r, prob[ZERO_CONTEXT_NODE])) {
INCREMENT_COUNT(ZERO_TOKEN);
dqv = dq[1];
token_cache[scan[c]] = 0;
++c;
if (c >= max_eob)
return c; // zero tokens at the end (no eob token)
ctx = get_coef_context(nb, token_cache, c);
band = *band_translate++;
prob = coef_probs[band][ctx];
}
// ONE_CONTEXT_NODE_0_
if (!vp9_read(r, prob[ONE_CONTEXT_NODE])) {
INCREMENT_COUNT(ONE_TOKEN);
WRITE_COEF_CONTINUE(1, ONE_TOKEN);
}
INCREMENT_COUNT(TWO_TOKEN);
prob = vp9_pareto8_full[prob[PIVOT_NODE] - 1];
if (!vp9_read(r, prob[LOW_VAL_CONTEXT_NODE])) {
if (!vp9_read(r, prob[TWO_CONTEXT_NODE])) {
WRITE_COEF_CONTINUE(2, TWO_TOKEN);
}
if (!vp9_read(r, prob[THREE_CONTEXT_NODE])) {
WRITE_COEF_CONTINUE(3, THREE_TOKEN);
}
WRITE_COEF_CONTINUE(4, FOUR_TOKEN);
}
if (!vp9_read(r, prob[HIGH_LOW_CONTEXT_NODE])) {
if (!vp9_read(r, prob[CAT_ONE_CONTEXT_NODE])) {
val = CAT1_MIN_VAL;
ADJUST_COEF(CAT1_PROB0, 0);
WRITE_COEF_CONTINUE(val, CATEGORY1_TOKEN);
}
val = CAT2_MIN_VAL;
ADJUST_COEF(CAT2_PROB1, 1);
ADJUST_COEF(CAT2_PROB0, 0);
WRITE_COEF_CONTINUE(val, CATEGORY2_TOKEN);
}
if (!vp9_read(r, prob[CAT_THREEFOUR_CONTEXT_NODE])) {
if (!vp9_read(r, prob[CAT_THREE_CONTEXT_NODE])) {
val = CAT3_MIN_VAL;
ADJUST_COEF(CAT3_PROB2, 2);
ADJUST_COEF(CAT3_PROB1, 1);
ADJUST_COEF(CAT3_PROB0, 0);
WRITE_COEF_CONTINUE(val, CATEGORY3_TOKEN);
}
val = CAT4_MIN_VAL;
ADJUST_COEF(CAT4_PROB3, 3);
ADJUST_COEF(CAT4_PROB2, 2);
ADJUST_COEF(CAT4_PROB1, 1);
ADJUST_COEF(CAT4_PROB0, 0);
WRITE_COEF_CONTINUE(val, CATEGORY4_TOKEN);
}
if (!vp9_read(r, prob[CAT_FIVE_CONTEXT_NODE])) {
val = CAT5_MIN_VAL;
ADJUST_COEF(CAT5_PROB4, 4);
ADJUST_COEF(CAT5_PROB3, 3);
ADJUST_COEF(CAT5_PROB2, 2);
ADJUST_COEF(CAT5_PROB1, 1);
ADJUST_COEF(CAT5_PROB0, 0);
WRITE_COEF_CONTINUE(val, CATEGORY5_TOKEN);
}
val = 0;
cat6 = cat6_prob;
while (*cat6)
val = (val << 1) | vp9_read(r, *cat6++);
val += CAT6_MIN_VAL;
WRITE_COEF_CONTINUE(val, CATEGORY6_TOKEN);
}
return c;
}
2010-05-18 19:58:33 +04:00
int vp9_decode_block_tokens(VP9_COMMON *cm, MACROBLOCKD *xd,
int plane, int block, BLOCK_SIZE plane_bsize,
int x, int y, TX_SIZE tx_size, vp9_reader *r) {
struct macroblockd_plane *const pd = &xd->plane[plane];
const int ctx = get_entropy_context(tx_size, pd->above_context + x,
pd->left_context + y);
const int eob = decode_coefs(cm, xd, block, pd->plane_type,
BLOCK_OFFSET(pd->dqcoeff, block), tx_size,
pd->dequant, ctx, r);
set_contexts(xd, pd, plane_bsize, tx_size, eob > 0, x, y);
return eob;
}