Allows DomainTxfm filters to be turned off for experimentation.
Also expands the parameter set for the Self guided filters.
Change-Id: I68fdb8e079a2464d80b3a4a990005c49baaaf0b8
Wedges were wrongly indexed when cb4x4 was on.
Brings a bit of the gain back when ext-inter + cb4x4 are turned on
together.
Change-Id: Id2bd359e70546cf0ea9cf31656064711c9894177
The reference frame handling in av1_find_mv_refs introduced
in https://aomedia-review.googlesource.com/c/6067/ broke the
compile when global-motion is enabled but ref-mv is not.
This patch adds the missing logic, allowing this case to compile
again.
Change-Id: I914887eb56d28a700b2917d086447bdbb314f35d
This commit makes the adaptive scan order system support multi-
thread encoding. It fixes unit test failure issue associated with
AV1/AVxEncoderThreadTest.EncoderResultTest/0.
BUG=aomedia:353
Change-Id: I61cbf9531c8deab97fb3bb17428d0b2a63cf309a
The WienerInfo struct requires a 16-byte alignment on x86,
since it contains filter coefficients which are loaded using
SSE aligned load instructions. But on 32-bit x86, the default
alignment of aom_malloc/aom_realloc is only 8 bytes, leading
to occasional segfaults.
To fix this, rather than using aom_realloc to resize WienerInfo
structures, we always free and re-allocate them using aom_memalign
BUG=aomedia:345
Change-Id: Ib1b2a42d4a2fa215dcc81ea481c51271ab068a37
CLPF performance had degraded by about 0.5% over the past six months,
which isn't totally surprising since the codec is a moving target.
About half of that degradation comes from the improved 7 bit filter
coefficients. Therefore, CLPF needs to be retuned for the current
codec.
This patch makes two (normative) changes to the CLPF kernel:
* The clipping function was changed from clamp(x, -s, s) to
sign(x) * max(0, abs(x) - max(0, abs(x) - s +
(abs(x) >> (bitdepth - 3 - log2(s)))))
This adds a rampdown to 0 at -32 and 32 (for 8 bit, -128 & 128
for 10 bit, etc), so large differences are ignored.
* 8 taps instead of 6 taps:
1
4 3
13 31 -> 13 31
4 3
1
AWCY results: low delay high delay
PSNR: -0.40% -0.47%
PSNR HVS: 0.00% -0.11%
SSIM: -0.31% -0.39%
CIEDE 2000: -0.22% -0.31%
APSNR: -0.40% -0.48%
MS SSIM: 0.01% -0.12%
About 3/4 of the gains come from the new clipping function.
Change-Id: Idad9dc4004e71a9c7ec81ba62ebd12fb76fb044a
The only compound mode used with sub 8x8 blocks is COMPOUND_AVERAGE, so
we don't have to send anything in this case
Change-Id: I90d0162e5f7f1ad205e65094293cde2a48eb77b1
When convolve_round is on, av1_convolve_2d_facade will be used for
interpolation rather than av1_convolve. Will remove the experiment
code of convolve_round experiment from av1_convolve in another CL.
So far we use 4-bit rounding in the intermediate stage on top of using
post rounding for compound mode after the last stage.
This will give us roughly 0.45% gain on lowres , 0.39% on midres and
roughly 0.6-0.7% on hdres
Altogether, is 1.15% on lowresm, 0.74% on midres and roughly 1.7-1.8% on
hdres
Note that there no restriction usage of 12-tap filter in the CL.
Adding that, we will lose roughly 0.1% again on lowres.
Change-Id: I6332e1d888e28a3b3ddc29711817d66e52cb5cdf
The new_tokenset experiment replaces the unconstrained tokenset with a
multisymbol alphabet in an inventive way.
Tested configurations:
new_tokenset + ec_adapt, new_tokenset, ec_multisymbol
Change-Id: I846ab2e51c2a1dc3f2f9904ed8c47a8e98f853c5
By default, the activity masking is used with PVQ.
In addition to '--enable-pvq', '--enable-daala-dist' is also
required by configure to use the activity masking.
Change-Id: I5100a1db992f0e693e61daf5439de8ae8c64a752
For fixed-point version of PVQ, which is current default,
added MAXI(1, ) to limit the minimum companded or expanded gain to be one.
Previously, gain compand/expand function, which is invoked when
activity masking is enabled, sometimes outputs zero
then triggered the assert(gain != 0).
Metric change from floating-pt to fixed-pt PVQ is:
PSNR PSNR-HVS SSIM CIEDE-2000 PSNR Cb PSNR Cr MS-SSIM VMAF
0.02 0.10 0.08 0.11 0.01 0.02 0.13 -0.30
Change-Id: I64a60d1970d35a26af227841e4a5e50a89ddc44c
Preparation for merging EOB_TOKEN. The block_zero value
corresponds to the first EOB_TOKEN: other EOB_TOKEN values will
be merged with non-zero values.
Change-Id: I94036783ee240fa916a79c544ecd716a9c24fa59
This commit renames deblocking_across_tiles to loopfilter_across_tiles,
to get ready for dering and clpf integration.
Change-Id: Id25b051da9b1e5cb92f35a9619662597462d9537
These are optimized for EOB_TOKEN being associated
with the current position, not the previous.
CBP tables cover EOB_TOKEN for the the whole block.
This change causes a performance regression until
EOB_TOKEN is merged into the coding scheme.
Change-Id: Ica3a12ed97285cbae204ce3cc1a7e658ebcacc9f
Allow the above combination of experiments to work together
correctly, fixing an encode/decode mismatch bug when they
were all enabled.
This change causes build_masked_compound(_highbd) to only
ever be called if CONFIG_SUPERTX is off, so wrap these functions
in an '#if !CONFIG_SUPERTX' block.
BUG=aomedia:313
Change-Id: Ic3886bc69ba9624b8fcb0a4c2d71fc64d2c0f22c
Zero, one, and two or more coded as one symbol (head).
Remaining tokens coded as a tail symbol.
The pareto CDF distribution is adjusted to cover tokens from
two onwards.
Change-Id: I98b33fab6b9f52690f6ad618ac55e725a97be056
- Added comments for some tables and #defines for clarity.
- Renamed some variables to ensure we use "color_index" instead of
"color" for palette color index related variables.
Change-Id: Ica95a26e0f171a41a3259c8e6b3b891b8cd10151
This commit makes the daala-ec work in the cb4x4 mode. As compared
to --enable-experimental, --enable-experimental --enable-cb4x4
improves the coding performance by:
lowres 2.6%
midres 1.2%
Change-Id: Ifee6f011c80364492c4a547513d24eb2958b5a56
Now that we have small number of contexts (5), use hash multipliers
(instead of base 11), so that color context hash is within a small
range. This allows us to use a lookup table to get color context
instead of a for loop.
Output bitstreams are bit-exact, so no change in compression.
Change-Id: I8cd8c893048c2fc6b22ccbd56f652d11486e2ee9
This reduces the complexity in a number of ways:
- We need just 3 neighbors instead of 4.
- Possible contexts reduce from 16 to 5.
- On hardware side, getting the contexts for a whole block will be more
parallelizable.
At the same time, compression performance improves very slightly:
- Screen-content set (videos) (Google): BDRate improved by 0.32
- screenshots set (images) (AWCY): PSNR improved by 0.62:
https://arewecompressedyet.com/?job=palette_withTR2%402017-01-27T21%3A30%3A28.890Z&job=palette_noTR2%402017-01-27T21%3A41%3A34.312Z
Change-Id: Ie84ca32f05d55ad481a51c2d3abc579468597189
This commit fixes the encoding/decoding mismatch issue when
ext-partition and ext-partition-type are both turned on in cb4x4
mode.
BUG=aomedia:336
Change-Id: I4d6ad5863c9d3bc8e3a41c259b8b39f130164790
Adjusts the value by 1 to make sure that the center tap
if the Wiener filter does not drop below 0.
BUG=aomedia:315
Change-Id: I41c3a2eb3f36dd49072a4873a995003d18f94ece
clang-format
Zoe Liu
David Barker
Debargha Mukherjee
Urvang Joshi
Tom Finegan
Angie Chiang
Jingning Han
Steinar Midtskogen
Thomas Daede
Sarah Parker
Alex Converse
David Michael Barr
Fangwen Fu
Yushin Cho
Thomas Davies
Yue Chen
Jonathan Matthews
Ryan Lei
Sebastien Alaiwan