Commit Graph

374 Commits

Author SHA1 Message Date
Jingning Han
696ee004a5 Re-work rate-distortion optimization scheme for transform coding
This commit re-works the rate-distortion optimization scheme for
transform coding. It improves the overall compression performance.
For derf set, the ext-tx experiment provides 2.27% coding gains,
and the new scheme that integrates multiple transform type selection
and recursive transform block partitioning provides a total of 3.24%
coding gains.

Change-Id: Ia1887c4c44b73dfb915d091d96660a99f09d5cc3
2015-11-03 09:03:53 -08:00
Jingning Han
4b594d3d00 Incorporate flexible tx type and tx partition in RD scheme
This commit hooks up the rate-distortion optimization system to
fully exploit recursive transform block partition and multiple
transform type. The compression performance of the two experiments
largely adds up. For derf set, ext-tx provides additional 2.1%
coding gains on top of the gains due to recursive transform block
partition (0.69%).

Change-Id: I1091fb9545f74e489a6a2489dc3c12f5abd05043
2015-11-02 17:40:05 -08:00
Jingning Han
94266f4f34 Make loop filter support recursive transform block partitioning
This commit allows the loop filter to account for the recursive
transform block partition when selecting the filter and mask.

Change-Id: I62b6c2dcc0497cbe1f264b03c46163f55d2c9752
2015-10-30 15:42:25 -07:00
Jingning Han
47c7fd984e Fix a switch condition in select_tx_block
Change-Id: I3d90a0286c5ef559b91ad298db97e8990becf85f
2015-10-30 13:01:52 -07:00
Jingning Han
bfeac5e19c Support per transform block skip coding
Allow the encoder to drop individual transform block coding.

Change-Id: I2c2b2985254cb92baf891f03daa33f067279373b
2015-10-30 08:55:17 -07:00
Jingning Han
3edad6e887 Enable entropy coding of recursive transform block partition
This commit enables the entropy coding of the recursive transform
block partition syntax.

Change-Id: I0c2509fb7b9822d12a721f9ebf9327fac83c777e
2015-10-29 11:06:46 -07:00
Debargha Mukherjee
8a4292441f Refactoring tx-types to add more flexibility
Allows inter and intra tx_types to have different sets of
transforms for different tx_size/sb_type combinations.

Change-Id: Ic0ac1daef7a9fb15c4210271e4d04cd36e5cec8e
2015-10-28 23:31:32 -07:00
Jingning Han
71c156070c Use precise distortion metric
Rework the rate distortion optimization pipeline. Use precise
distortion metric that accounts for the forward and inverse
transform rounding effect.

Change-Id: Ibe19ce9791ec3547739294cc3012dd9e11f4ea49
2015-10-28 11:47:14 -07:00
Jingning Han
4bfed0b32e Account for variable txfm sizes in coeff token packing
This commit makes the coefficient token packtization process account
for variable transform block sizes supported in a single processing
block. It fixes an enc/dec mismatch issue when var-tx, ext-tx, and
misc-fixes experiments are all turned on.

Change-Id: I2e8946e6f72de567603a568debbadad11196430c
2015-10-28 11:45:31 -07:00
Yaowu Xu
b6da40ad82 Merge branch 'master' into nextgenv2
Change-Id: I0e4030a37354bb23b3aa8be5cc1473770b9e7b06
2015-10-27 08:28:09 -07:00
Jingning Han
236623cf2c Fix early termination flag in recursive transform block search
Properly reset the early termination flag in the recursive transform
block partitioning rate-distortion optimization scheme.

Change-Id: Ibfe918f21f11dcb1ec267c09f954c635305cc95a
2015-10-26 19:41:14 -07:00
Jingning Han
f0dee7765a Fix lossless coding
Use inter_block_yrd as rate-distortion optimization for lossless
coding. This fixes transform coefficient buffer swap use case and
resolves the unit test failure related to lossless coding.

Change-Id: I1512dab5ed5760c31f7de21a06e8d9ed1eb081fa
2015-10-26 22:03:39 +00:00
Jingning Han
79fe7246c1 Properly handle non-420 color format in recursive transform scheme
This commit makes the recursive transform block partitioning properly
handle the non-420 color format. It resolves an enc/dec mismatch
issue in that setting when var-tx experiment is turned on.

Change-Id: I48a91de02c11b3153f897d1cca0ae948eec15605
2015-10-23 14:42:01 -07:00
Jingning Han
caeb10bf06 Use explicit block position in foreach_transformed_block
Add the row and column index to the argument list of unit functions
called by foreach_transformed_block wrapper. This avoids the
repeated internal parsing according to the block index.

Change-Id: Ie7508acdac0b498487564639bc5cc6378a8a0df7
2015-10-23 09:19:17 -07:00
Yaowu Xu
df33be4b22 Merge "Fix merge defects" into nextgenv2 2015-10-23 01:28:32 +00:00
Yaowu Xu
5a27b3bb85 Fix merge defects
This commit fixes the merge conflicts between master and nextgenv2 and
disable early termination in choose_tx_size() to avoid failure in test.

The test failures are pre-existing, some of the issue were fixed in
masterbase already, so will have another merge to introduce the fixes.

Change-Id: Ib71889661955e73aedbb4db49d8be70425281dcb
2015-10-22 18:25:41 -07:00
Jingning Han
18aea429af Merge "Reset tx_type in recursive transform block partitioning" into nextgenv2 2015-10-22 21:34:32 +00:00
Jingning Han
d3e5545fa6 Reset tx_type in recursive transform block partitioning
Temporarily reset the transform type in the inter modes when
recursive transform block partitioning is used. This resolves an
enc/dec mismatch issue in nextgenv2 codebase when both var-tx and
ext-tx experiments are turned on.

Change-Id: I2543f0a567243da95b237752d46964b07b669ad9
2015-10-22 14:28:56 -07:00
Yaowu Xu
4ac2ae3a4d Merge branch 'masterbase' into nextgenv2
Conflicts:
	configure
	test/vp9_encoder_parms_get_to_decoder.cc
	vp10/common/blockd.h
	vp10/common/entropymode.c
	vp10/common/entropymode.h
	vp10/common/idct.c
	vp10/decoder/decodeframe.c
	vp10/decoder/decodemv.c
	vp10/encoder/bitstream.c
	vp10/encoder/encodeframe.c
	vp10/encoder/encodemb.c
	vp10/encoder/encoder.c
	vp10/encoder/encoder.h
	vp10/encoder/rd.c
	vp10/encoder/rdopt.c
	vp10/encoder/tokenize.c
	vp10/encoder/tokenize.h
	vp9/decoder/vp9_decodeframe.c
	vp9/decoder/vp9_decoder.h
	vp9/encoder/vp9_aq_cyclicrefresh.c
	vp9/encoder/vp9_encoder.h
	vp9/vp9_cx_iface.c
	vpx/vp8cx.h
	vpx_dsp/x86/vpx_subpixel_8t_intrin_ssse3.c
	vpx_scale/yv12config.h

Change-Id: I604a329d38badec7a11e8ede16ca1404476e9b93
2015-10-22 11:40:44 -07:00
Jingning Han
20484048a9 Fix compiler error with high bit-depth and var-tx
Clear the compiler errors when both high bit-depth and recursive
transform block partition experiments are enabled.

Change-Id: If0b6396851f10c28b4f26350322ccd1ba2fc9aff
2015-10-21 17:38:00 -07:00
Hui Su
96b69deca5 Merge "VP10: some changes to palette mode" 2015-10-20 16:37:31 +00:00
Ronald S. Bultje
60c58b5284 vp10: per-segment lossless coding.
Some more testing of this patch would probably be useful, but I
think the basics of it should work fine now.

See issue 1035.

Change-Id: I4a36d58f671c5391cb09d564581784a00ed26245
2015-10-16 19:30:39 -04:00
Ronald S. Bultje
c7dc1d78bf vp10: add extended-intra prediction edges experiment.
This experiment allows using full above/right edges for all transform
sizes whenever available (for d45/d63), and adds bottom/left edges for
d207.

See issue 1043.

Change-Id: I5cf7f345e783e8539bb6b6d2c9972fb1d6d0a78b
2015-10-16 19:30:39 -04:00
Ronald S. Bultje
d8f3bb1837 vp10: merge keyframe/interframe uvintramode/partition probabilities.
This has various benefits:
- simplify implementations because we don't have to switch between
  multiple probability tables depending on frametype
- allows fw subexp and bw adaptivity for partitions/uvmode in keyframes

See issue 1040 point 5.

Change-Id: Ia566aa2863252d130cee9deedcf123bb2a0d3765
2015-10-16 19:30:38 -04:00
hui su
17c817adfc VP10: some changes to palette mode
Account for rounding in distortion calculation in k-means;
carry out rounding before duplicates removal of base colors;
replace numbers with macros;
use prefix increment.

Slight coding gain (<0.1%) on screen_content testset.

Change-Id: Ie8bd241266da6b82c7b2874befc3a0c72b4fcd8c
2015-10-16 11:41:26 -07:00
hui su
aaf6f6215f Fix palette mode in multi-thread encoding setting
Fix a couple of memory related errors. Also fix thread test failures.

Change-Id: I0103995f832cecf1dd2380000321ac7204f0cfc0
2015-10-15 15:00:57 -07:00
Yaowu Xu
15cc8bc72f Merge "fix a msvc compiler warning" 2015-10-15 14:39:01 +00:00
Yaowu Xu
8ced62f250 fix a msvc compiler warning
Change-Id: Ifd6581c1bdb8d8f4b2ecf676c1a3d385dc129abf
2015-10-15 01:05:13 +00:00
Yaowu Xu
4727fa2a75 Fix two asan failures
Change-Id: I57865e9604ac162ef0d97deb16e81ca436a98428
2015-10-14 18:03:31 -07:00
Jingning Han
1e48f74d9a Enable early termination in the recursive transform block search
It makes the encoder 5% faster for CIF clips and 12% faster for
1080p clips.

Change-Id: I073408dbb4d50675a79db8794fe73975ac957b91
2015-10-13 21:14:58 +00:00
Jingning Han
3a27961cf3 Refactor recursive transform block scheme
This commit re-designs the recursive transform block partition
rate-distortion optimization framework. It allows the encoder to
improve speed by 10%.

Change-Id: I6dd3a7dd428a530d8012e5c6ddc40e650c8b392b
2015-10-13 21:13:29 +00:00
Jingning Han
2cdc12742d Rate-distortion optimization for recursive transform block coding
This commit enables the rate-distortion optimization for recursive
transform block coding scheme.

Change-Id: Id6a8336ca847bb3af1e94cbfb51db1f4da12d38f
2015-10-13 12:49:03 -07:00
Hui Su
b9e31b5163 Merge "VP10: Add palette mode part 1" 2015-10-13 17:34:27 +00:00
Ronald S. Bultje
5b4805d6e9 vp10: remove clamp_mv2() call from vp10_find_best_ref_mvs().
This actually has no effect whatsoever, since the input MVs themselves
are clamped by clamp_mv_ref() already, which is significantly more
restrictive in its bounds.

Change-Id: I4a3a7b2b121ee422c56428c2a12d930c3813c06e
2015-10-12 14:45:18 -04:00
hui su
5d011cb278 VP10: Add palette mode part 1
Add palette mode for keyframe luma channel. Palette mode is enabled
when using "--tune-content=screen" in encoding config parameters.

on screen_content testset:  +6.89%
on derlr                 :  +0.00%

Design doc (WIP):
https://goo.gl/lD4yJw

Change-Id: Ib368b216bfd3ea21c6c27436934ad87afdaa6f88
2015-10-12 10:02:17 -07:00
Jingning Han
a8dad55c82 Make chroma component RD estimate support transform partition
This commit makes the rate-distortion optimization for chroma
component support the recursive transform block coding scheme.

Change-Id: I1bfed6d05b0ebb3905cb625222401e2ccbae10f3
2015-10-08 18:04:03 -07:00
Jingning Han
704985e65a Add decoder support to recursive transform block partition
This commit allows the decoder to recursively parse and rebuild
the pixel blocks.

Change-Id: I510f3a30ae7cdad5b70725c66882b00a0594e96f
2015-10-08 16:16:41 -07:00
Jingning Han
ebc48efe37 Use explicit block position in foreach_transformed_block
Add the row and column index to the argument list of unit functions
called by foreach_transformed_block wrapper. This avoids the
repeated internal parsing according to the block index.

Change-Id: I42b3578eac258ebaba7a7c74f684de9abab521a6
2015-10-07 16:32:19 -07:00
Jingning Han
00ca5c1c98 Simplify vp10_xform_quant index parsing
Change-Id: Id7f7a9b2e53fc0074b55d58143f296afad6b844e
2015-10-06 17:19:23 -07:00
hui su
4f16f11993 Refactor ext-tx experiment
Remove unnecessary transform type lookups and unused codes.

Change-Id: Ib52d26690468996b1501b419d919643c8ea5ecaa
2015-10-02 11:00:51 -07:00
hui su
3fa0129caf Extend ext_tx experiment to intra blocks
ext-tx on derflr +2.30% (was +1.84%)

Change-Id: Ic91565cacc38e7a8e1200d054ed7bf99295fe19e
2015-10-02 09:39:38 -07:00
hui su
2afe7320c8 Add identity transform to ext-tx experiment
ext-tx on derflr: +1.756% (was +1.648)

Change-Id: I8a87970fa589e8f5f96db7aa68ec9b6c98e20188
2015-09-30 18:47:46 -07:00
Hui Su
4b7043f804 Merge "ext-tx experiment support in choose_largest_tx_size" into nextgenv2 2015-09-30 15:14:17 +00:00
Debargha Mukherjee
3e8cceb3fc Speed up of DST and the search in ext_tx
Adds an early termination to the ext_tx search, and also
implements the DST transforms more efficiently.

About 4 times faster with the ext-tx experiment.

There is a 0.09% drop in performance on derflr from 1.735% to
1.648%, but worth it with the speedup achieved.

Change-Id: I2ede9d69c557f25e0a76cd5d701cc0e36e825c7c
2015-09-29 19:11:43 -07:00
Yaowu Xu
7c514e2dfd Merged branch 'master' into nextgenv2
Resolved Conflicts in the following files:
        configure
        vp10/common/idct.c
        vp10/encoder/dct.c
        vp10/encoder/encodemb.c
        vp10/encoder/rdopt.c

Change-Id: I4cb3986b0b80de65c722ca29d53a0a57f5a94316
2015-09-29 16:17:32 -07:00
hui su
6c81e37916 ext-tx experiment support in choose_largest_tx_size
Change-Id: Ic161b8b257a02c1c43e515d830c1051d0de074de
2015-09-29 12:09:15 -07:00
Ronald S. Bultje
bab8d38f7f vp10: remove MACROBLOCK.{highbd_,}itxfm_add function pointer.
This is preparatory work for allowing per-segment lossless coding.

See issue 1035.

Change-Id: I9487d02717ee3e766aee61a487780056bb35d2d3
2015-09-25 19:30:46 -04:00
Ronald S. Bultje
c74b33a413 vp10: remove MACROBLOCK.fwd_txm4x4 function pointer.
This is preparatory work for allowing per-segment lossless coding.

See issue 1035.

Change-Id: Idd72e2a42d90fa7319c10122032d1a7c7a54dc05
2015-09-25 19:30:46 -04:00
hui su
38cc168822 Adjust rd calculation in choose_tx_size_from_rd
Coding gain:
derflr 0.142%
hevclr 0.153%
hevcmr 0.124%

Change-Id: I63b56ae3a9002c3a266e10e2964135ed43b0ba53
2015-09-23 10:54:28 -07:00
hui su
07154b0216 Refactor ext-tx experiment
At this point, ext-tx compapred to baseline +1.735%.

Change-Id: Ia16ac293e2cc87e06a0d898c1d52a8f3495ff814
2015-09-23 09:14:49 -07:00