]> granicus.if.org Git - libvpx/log
libvpx
9 years agoAdded more LAST2 modes for INTERINTRA
Zoe Liu [Fri, 11 Sep 2015 23:55:24 +0000 (16:55 -0700)]
Added more LAST2 modes for INTERINTRA

Turning on all the other experiments, compared the RD performance
between with and without the use of LAST2_FRAME, on derflr testset,
on Average PSNR:

8-bit: +0.653% (All positive except one,
max: mobile_cif: 2.019%; min: paris_cif: -0.081%)
12-bit HBD: +0.735% (All positive,
max: bridge_far_cif: 2.416%; min: bowing_cif: 0.132%)

Change-Id: Ia0a375667e228c8ba3d2e223abff608206f2f545

9 years agoAdded MACRO for reference frame encoding
Zoe Liu [Fri, 11 Sep 2015 21:57:31 +0000 (14:57 -0700)]
Added MACRO for reference frame encoding

This CL introduces a few macros plus code cleaning on the encoding of
the reference frames. Coding performance remains unchanged.

For the encoding of either the compound reference or the single reference
case, since each bit has different contexts, the tree structure may not
be applied to treat the combined bits as one symbol. It is possible we may
explore the sharing of the same context for all the bits to introduce
the use of tree structure for the next step.

Change-Id: I6916ae53c66be1a0b23e6273811c0139515484df

9 years agoAdded one more reference frame LAST2_FRAME
Zoe Liu [Fri, 21 Aug 2015 02:08:46 +0000 (19:08 -0700)]
Added one more reference frame LAST2_FRAME

Under the experiment CONFIG_MULTI_REF. Current version shows
LAST2 vs base in nextgen on the testset of derflr:

(1) 8-bit: Average PSNR +0.53%
(worst: students_cif: -0.247%; best: mobile_cif: 1.902%)
(2) 12-bit HBD: Average PSNR +0.63%
(worst: pamphlet_cif: -0.213%, best: mobile_cif: 2.101%)

More tuning on the reference frame context design and default
probs are being conducted. This version does not guarantee to
work with other experiments in nextgen. A separate CL will address
the working with all other experiments.

Change-Id: I7f40d2522517afc26ca389c995bad56989587f65

9 years agoSuper resolution mode (+CONFIG_SR_MODE)
Shunyao Li [Wed, 12 Aug 2015 01:29:41 +0000 (18:29 -0700)]
Super resolution mode (+CONFIG_SR_MODE)

CONFIG_SR_MODE=1, enable SR mode
USE_POST_F=1, enable SR post filter
SR_USE_MULTI_F=1, enable SR post filter family
Not compatible with other experiments yet

Change-Id: I116f1d898cc2ff7dd114d7379664304907afe0ec

9 years agoAdd transform size rate for intra skip mode in rdopt
Shunyao Li [Fri, 28 Aug 2015 22:08:06 +0000 (15:08 -0700)]
Add transform size rate for intra skip mode in rdopt

Change-Id: I81fedd99cd39c12b66b93b786cb43234c867b84b

9 years agoMake tests work with new configurations
Debargha Mukherjee [Thu, 27 Aug 2015 19:28:59 +0000 (12:28 -0700)]
Make tests work with new configurations

Disables some test vector tests when Vp8/Vp9 decoders are disabled
in configuration. Also moves some macros to the vpx level in
line with recent refactoring on the master branch.

Change-Id: Iaac8008992110398ae096c36b9726f723164c207

9 years agoSome tweaks to probabilities for ext-tx with dst1
Debargha Mukherjee [Wed, 12 Aug 2015 21:31:31 +0000 (14:31 -0700)]
Some tweaks to probabilities for ext-tx with dst1

derflr: up to 1.429% from a little less than 1.3% when
--enable-dst1 is also enabled with --enable-ext-tx.

Change-Id: I301229f2239b18acb96accc4fc44b64fa6927ace

9 years agocode cleanup in encode_block_intra
hui su [Thu, 30 Jul 2015 18:31:59 +0000 (11:31 -0700)]
code cleanup in encode_block_intra

Change-Id: I376b7e9b243178d79141a96e0aeafcbc15758e97

9 years agoResolve bug of DST1 in ext_tx experiment.
Cheng Chen [Tue, 28 Jul 2015 17:49:55 +0000 (10:49 -0700)]
Resolve bug of DST1 in ext_tx experiment.

Change-Id: I828569e3596f9b9e8487aec7c4056e66cf1fc1f2

9 years agoAdds support for DST1 transforms for inter blocks
Debargha Mukherjee [Thu, 23 Jul 2015 20:14:39 +0000 (13:14 -0700)]
Adds support for DST1 transforms for inter blocks

Adds an additional transform in the ext_tx experiment that
is a 2d DST1-DST1 combination.

To enable use --enable-ext-tx --enable-dst1.

This needs to be later extended to combine DST1 with DCT
or ADST.

Change-Id: I6d29f1b778ef8294bcfb6a512a78fc5eda20723b

9 years agoSpeed up of supertx
Shunyao Li [Fri, 17 Jul 2015 17:44:06 +0000 (10:44 -0700)]
Speed up of supertx

Limited the prediction extension to 8 pixels at each edge
Fixed a bug in the combination of wedge prediction and supertx

~10% speed up in decoder
derflr:     -0.004
derflr+hbd: +0.002
hevcmr:     +0.015

Change-Id: I777518896894a612c9704d3de0e7902bf498b0ea

9 years agoAdd extended transforms for 32x32 and 64x64
Debargha Mukherjee [Fri, 17 Jul 2015 05:13:40 +0000 (22:13 -0700)]
Add extended transforms for 32x32 and 64x64

Framework for alternate transforms for inter 32x32 and larger based
on dwt-dct hybrid is implemented.
Further experiments are to be condcuted with different
variations of hybrid dct/dwt or plain dwt, as well as super-resolution
mode.

Change-Id: I9a2bf49ba317e7668002cf1499211d7da6fa14ad

9 years agoNextgen branch cleanup: remove elevate_newmv_thresh feature
Yunqing Wang [Tue, 14 Jul 2015 21:28:56 +0000 (14:28 -0700)]
Nextgen branch cleanup: remove elevate_newmv_thresh feature

Code cleanup.

Change-Id: Idf00d5359e5f3d943ee39b4a00f8d40325c0f8b3

9 years agoNextgen branch cleanup: remove real-time speed features
Yunqing Wang [Tue, 14 Jul 2015 21:09:35 +0000 (14:09 -0700)]
Nextgen branch cleanup: remove real-time speed features

Removed inter_mode_mask and max_intra_bsize.

Change-Id: I835d31f95af64a524b2fdb685d1d09a6b0ad5c9f

9 years agoMerge "Nextgen branch cleanup: add back some SVC functions" into nextgen
Yunqing Wang [Wed, 8 Jul 2015 23:26:33 +0000 (23:26 +0000)]
Merge "Nextgen branch cleanup: add back some SVC functions" into nextgen

9 years agoNextgen branch cleanup: add back some SVC functions
Yunqing Wang [Wed, 8 Jul 2015 20:37:33 +0000 (13:37 -0700)]
Nextgen branch cleanup: add back some SVC functions

Several SVC functions were added back to ensure Borg test build.

Change-Id: I2cff676407fa74e2255606094ff12bd840287b6d

9 years agoA small code cleanup for the encoder workflow.
Zoe Liu [Wed, 8 Jul 2015 19:12:47 +0000 (12:12 -0700)]
A small code cleanup for the encoder workflow.

Change-Id: I543a69ad5a34cd9a0671b57131cafc541d595d9a

9 years agoMerge "intrabc: Allow odd pel displacement for non-444" into nextgen
Alex Converse [Wed, 8 Jul 2015 04:06:23 +0000 (04:06 +0000)]
Merge "intrabc: Allow odd pel displacement for non-444" into nextgen

9 years agoMerge "Optimize bilateral filter to improve speed" into nextgen
Shunyao Li [Tue, 7 Jul 2015 20:17:12 +0000 (20:17 +0000)]
Merge "Optimize bilateral filter to improve speed" into nextgen

9 years agoNextgen branch cleanup: remove svc code
Yunqing Wang [Mon, 6 Jul 2015 21:39:03 +0000 (14:39 -0700)]
Nextgen branch cleanup: remove svc code

Spatial/temporal svc code was removed.

Change-Id: Ie25c7a58ee5feb662d4de526406d8cd834d19977

9 years agoNextgen branch cleanup: remove unused functions
Yunqing Wang [Mon, 6 Jul 2015 19:05:26 +0000 (12:05 -0700)]
Nextgen branch cleanup: remove unused functions

Removed unused real-time functions.

Change-Id: I0cbcee67fb8f495f87c6330e04ed6d56ed8e2625

9 years agoNextgen branch cleanup: remove nonrd code
Yunqing Wang [Mon, 6 Jul 2015 17:03:23 +0000 (10:03 -0700)]
Nextgen branch cleanup: remove nonrd code

Code cleanup.

Change-Id: I4d8d7fa2fc61a58b819c9a18bf25cda53e3fd88c

9 years agointrabc: Allow odd pel displacement for non-444
Alex Converse [Wed, 24 Jun 2015 18:17:50 +0000 (11:17 -0700)]
intrabc: Allow odd pel displacement for non-444

intrabc:
screen_content: -0.618 derflr: +0.015

intrabc+tx_skip+palette:
screen_content: -0.124 derflr: -0.048

Change-Id: Iabea4be19dce2f6fdab8f639e585a424c90c81b4

9 years agoOptimize bilateral filter to improve speed
Shunyao Li [Mon, 29 Jun 2015 18:54:17 +0000 (11:54 -0700)]
Optimize bilateral filter to improve speed

Optimization of bilateral filter:
1) Pre-calculate the bilateral filters at all the
levels at the initialization.
2) Convert 1D matrix to 2D matrix, avoid too many
multiplications in the bilateral filter loop.
3) Fix a bug in "loop_bilateral_filter_highbd".
The right-shifted range index can be larger than 255.

Change-Id: I42f6177e896706948e2403bd9edce46e3eb5cbf8

9 years agoMerge "intrabc: displacement vector prediction" into nextgen
Alex Converse [Thu, 25 Jun 2015 20:53:51 +0000 (20:53 +0000)]
Merge "intrabc: displacement vector prediction" into nextgen

9 years agointrabc: displacement vector prediction
Alex Converse [Mon, 27 Apr 2015 16:40:48 +0000 (09:40 -0700)]
intrabc: displacement vector prediction

Predict displacement vector with the same logic as NEARESTMV. If no
neighbors are available fall back to the current one block left or up
prediction.

vp9+intrabc+tx_skip+palette: -0.489
vp9+intrabc: -0.771

Change-Id: If67d08b54f1a3b847cf7ab8c7b800c55baa1a86b

9 years agoFix for copy-mode with segmentation
Peter de Rivaz [Wed, 24 Jun 2015 08:18:36 +0000 (09:18 +0100)]
Fix for copy-mode with segmentation

Change-Id: Icfc97826e494b856aa80fbefe4811dff99f3a0b4

9 years agoMerge "Removing some compiler warnings" into nextgen
Debargha Mukherjee [Tue, 23 Jun 2015 19:13:08 +0000 (19:13 +0000)]
Merge "Removing some compiler warnings" into nextgen

9 years agoAdded support for extended partitions with copy_mode
Peter de Rivaz [Tue, 23 Jun 2015 11:00:05 +0000 (12:00 +0100)]
Added support for extended partitions with copy_mode

Change-Id: I815374bd89e3faf2bac1b4155dbbe7366f0a1d29

9 years agoMerge "Fix a bug induced by qctx-tprobs experiment" into nextgen
hui su [Tue, 23 Jun 2015 17:54:35 +0000 (17:54 +0000)]
Merge "Fix a bug induced by qctx-tprobs experiment" into nextgen

9 years agoMerge "Bug fix in tx-skip" into nextgen
hui su [Tue, 23 Jun 2015 17:42:41 +0000 (17:42 +0000)]
Merge "Bug fix in tx-skip" into nextgen

9 years agoFix a bug induced by qctx-tprobs experiment
hui su [Tue, 23 Jun 2015 17:20:27 +0000 (10:20 -0700)]
Fix a bug induced by qctx-tprobs experiment

Change-Id: I8d31b8e6037464c36b514fddccd5d7edee024a47

9 years agoRemoving some compiler warnings
Debargha Mukherjee [Mon, 22 Jun 2015 22:26:25 +0000 (15:26 -0700)]
Removing some compiler warnings

Change-Id: I572594a9c00ba5a2b8000d3b4c86ec5757aa586c

9 years agoAdded palette support to extended partitions
Peter de Rivaz [Mon, 22 Jun 2015 19:34:26 +0000 (20:34 +0100)]
Added palette support to extended partitions

Change-Id: Ide11bd3fcc9fb26b27ae39cdf216a2c32b853348

9 years agoPalette high bit depth functionality
Julia Robson [Tue, 2 Jun 2015 13:31:38 +0000 (14:31 +0100)]
Palette high bit depth functionality

Changes to allow high bit depth and palette to be enabled at the
same time by using a 16 bit (instead of 8bit) palette when high
bit depth is enabled and modifying related functions accordingly.

Change-Id: I97d30b4d9338d3a51db02c94bc568eba60a8905d

9 years agoMerge "Add q-index as context for initial token probs" into nextgen
hui su [Mon, 22 Jun 2015 17:44:25 +0000 (17:44 +0000)]
Merge "Add q-index as context for initial token probs" into nextgen

9 years agoMerge "Fix for supertx decode with segments" into nextgen
Debargha Mukherjee [Mon, 22 Jun 2015 17:28:30 +0000 (17:28 +0000)]
Merge "Fix for supertx decode with segments" into nextgen

9 years agoMerge "Fix supertx HORZ and VERT partitioning." into nextgen
Debargha Mukherjee [Mon, 22 Jun 2015 17:11:38 +0000 (17:11 +0000)]
Merge "Fix supertx HORZ and VERT partitioning." into nextgen

9 years agoBug fix in tx-skip
hui su [Mon, 22 Jun 2015 16:54:42 +0000 (09:54 -0700)]
Bug fix in tx-skip

This patch avoids using tx_size larger than 16x16 in lossless mode.
Big block quantization (32x32 or larger) is not lossless.

Change-Id: I69cd84d4f3fd06d641048d6096da1bfde18ad24e

9 years agoAdd q-index as context for initial token probs
hui su [Sun, 31 May 2015 16:41:12 +0000 (09:41 -0700)]
Add q-index as context for initial token probs

There are 4 entropy tables to select for initial entropy table,
depending on the frame base q-index. The entropy tables are
trained with derf, yt, and stdhd sets. About 0.2% gain on
the following test sets:

derflr       0.227%
yt           0.277%
stdhd        0.233%
hevclr       0.221%
hevcmr       0.155%
hevchr       0.182%

Change-Id: I3fde846c47fc020e80c814897690b4cda1da569c

Change-Id: I460408372586c823974f945ed9fd8dcb0360fbaf

9 years agoFix for supertx decode with segments
Peter de Rivaz [Thu, 18 Jun 2015 18:15:15 +0000 (19:15 +0100)]
Fix for supertx decode with segments

Change-Id: I5e07b0b9fafd2578ecfacf20fb70b07fd6977d98

9 years agoFix supertx HORZ and VERT partitioning.
Peter de Rivaz [Wed, 17 Jun 2015 20:19:39 +0000 (21:19 +0100)]
Fix supertx HORZ and VERT partitioning.

Change-Id: I9a44d70dc2514baf679025335badc2744eeb4b88

9 years agoAdded ext_partitions to supertx
Peter de Rivaz [Wed, 17 Jun 2015 08:41:15 +0000 (09:41 +0100)]
Added ext_partitions to supertx

Change-Id: I0e2a0253e5a8dcdb5b765836a3301299409e9d0a

9 years agoFix seg fault in count_segs_sb with ext_partitions
Peter de Rivaz [Mon, 15 Jun 2015 10:03:38 +0000 (11:03 +0100)]
Fix seg fault in count_segs_sb with ext_partitions

Change-Id: I609f429006e7d2b849fe0a509f5d9f5123ecf0ec

9 years agoFixing skip simplifcation with SuperTX experiment
Pieter Kapsenberg [Mon, 15 Jun 2015 21:51:16 +0000 (14:51 -0700)]
Fixing skip simplifcation with SuperTX experiment

Combination of MISC_ENTROPY and SUPERTX resulted in compilation error

Change-Id: Ibcf28216712d93f00f38f3786fce8fcdca3f835e

9 years agoMerge "Fix tx-skip warnings on Windows" into nextgen
hui su [Mon, 15 Jun 2015 19:41:23 +0000 (19:41 +0000)]
Merge "Fix tx-skip warnings on Windows" into nextgen

9 years agoMerge changes I9d80152b,I4e13da4b into nextgen
Alex Converse [Mon, 15 Jun 2015 16:57:05 +0000 (16:57 +0000)]
Merge changes I9d80152b,I4e13da4b into nextgen

* changes:
  intrabc: Add a frame level flag
  intrabc: Remove NEWDV from the intra mode tree.

9 years agoMerge "Added tests for high bit depth quantize functions" into nextgen
Debargha Mukherjee [Fri, 12 Jun 2015 20:59:15 +0000 (20:59 +0000)]
Merge "Added tests for high bit depth quantize functions" into nextgen

9 years agoBug fix in highbd psnr computation with postproc
Debargha Mukherjee [Fri, 12 Jun 2015 18:56:42 +0000 (11:56 -0700)]
Bug fix in highbd psnr computation with postproc

Change-Id: I6944a5fc38d6a89e1bb739938c6ecae7ec731b8d

9 years agoAllow extended partition support to encode
Peter de Rivaz [Wed, 10 Jun 2015 11:19:59 +0000 (12:19 +0100)]
Allow extended partition support to encode

Change-Id: I75246e2ee35a1b7c1ad46669c464e582e3a9961c

9 years agoMerge "Add extended partition support to decode" into nextgen
Debargha Mukherjee [Thu, 11 Jun 2015 18:47:28 +0000 (18:47 +0000)]
Merge "Add extended partition support to decode" into nextgen

9 years agoFix tx-skip warnings on Windows
hui su [Thu, 11 Jun 2015 18:20:16 +0000 (11:20 -0700)]
Fix tx-skip warnings on Windows

Change-Id: I2a515abe20e6989de233fc8ae2c31d8ee462add2

9 years agoMerge "Fix palette build on windows" into nextgen
hui su [Thu, 11 Jun 2015 17:56:26 +0000 (17:56 +0000)]
Merge "Fix palette build on windows" into nextgen

9 years agoFix palette build on windows
hui su [Thu, 11 Jun 2015 17:11:45 +0000 (10:11 -0700)]
Fix palette build on windows

Change-Id: I2a90e235ab5bdf95eb83f684e46ef844553ba629

9 years agoAdd extended partition support to decode
Peter de Rivaz [Tue, 9 Jun 2015 10:50:35 +0000 (11:50 +0100)]
Add extended partition support to decode

Change-Id: I65f7aed37d65dbba76999d2b9585129c48e70b11

9 years agoAdded extended partition experiment
Peter de Rivaz [Tue, 9 Jun 2015 08:00:14 +0000 (09:00 +0100)]
Added extended partition experiment

Change-Id: Ife89c95f04212e9dceb4c4e663280a85b5718c13

9 years agoMerge "Fix copy mode bug: searches made outside of the current tile" into nextgen
Debargha Mukherjee [Thu, 11 Jun 2015 16:30:22 +0000 (16:30 +0000)]
Merge "Fix copy mode bug: searches made outside of the current tile" into nextgen

9 years agoFix copy mode bug: searches made outside of the current tile
Jack Haughton [Tue, 9 Jun 2015 13:34:14 +0000 (14:34 +0100)]
Fix copy mode bug: searches made outside of the current tile

Change-Id: Ib3cf4bf4da0a575053493609956e133d1ce028a1

9 years agoMerge "Fix to loopfilter tests" into nextgen
Debargha Mukherjee [Thu, 11 Jun 2015 05:30:49 +0000 (05:30 +0000)]
Merge "Fix to loopfilter tests" into nextgen

9 years agoMerge "Fix cost array overrun bug in search_bilateral_level()" into nextgen
Debargha Mukherjee [Thu, 11 Jun 2015 05:28:42 +0000 (05:28 +0000)]
Merge "Fix cost array overrun bug in search_bilateral_level()" into nextgen

9 years agointrabc: Add a frame level flag
Alex Converse [Wed, 20 May 2015 00:47:40 +0000 (17:47 -0700)]
intrabc: Add a frame level flag

For the combination of this and removing NEWDV from the tree:
derflr: -0.101 screen_content: +0.053

The bulk of the decline in screen content effecincy is from the liquify
clip. These should be recoverable by further entropy tweaks.

Change-Id: I9d80152b8492e60a0367c31797fb6932fb09bba9

9 years agoMerge "Fix masked_variance overflow for non-highbitdepth" into nextgen
Debargha Mukherjee [Tue, 9 Jun 2015 17:07:12 +0000 (17:07 +0000)]
Merge "Fix masked_variance overflow for non-highbitdepth" into nextgen

9 years agoFix cost array overrun bug in search_bilateral_level()
Jack Haughton [Tue, 9 Jun 2015 13:21:55 +0000 (14:21 +0100)]
Fix cost array overrun bug in search_bilateral_level()

Change-Id: I9b7b727cb45328040a82477a2ce8eb3d15feec77

9 years agoFix to loopfilter tests
Julia Robson [Tue, 9 Jun 2015 12:09:09 +0000 (13:09 +0100)]
Fix to loopfilter tests

Ensuring the random numbers used by the loopfilter tests
are in a valid range

Change-Id: If8130145c413dbe7e29180046eca3e4f7d941706

9 years agoAdds wavelet transforms + hybrid dct/dwt variants
Debargha Mukherjee [Tue, 2 Jun 2015 19:25:28 +0000 (12:25 -0700)]
Adds wavelet transforms + hybrid dct/dwt variants

The wavelets implemented are 2/6, 5/3 and 9/7 each with
a lifting based scheme for even block sizes. The 9/7
one is a double implementation currently.

This is to start experiments with:
1. Replacing large transforms (32x32 and 64x64) with wavelets
or wavelet-dct hybrids that can hopefully localize errors better
spatially. (Will also need alternate entropy coder)
2. Super-resolution modes where the higher sub-bands may be
selectively skipped from being conveyed, while a smart
reconstruction recovers the lost frequencies.

The current patch includes two types of 32x32 and 64x64
transforms: one where only wavelets are used, and another
where a single level wavelet decomposition is followed
by a lower resolution dct on the low-low band.

Change-Id: I2d6755c4e6c8ec9386a04633dacbe0de3b0043ec

9 years agoFix masked_variance overflow for non-highbitdepth
Jack Haughton [Mon, 8 Jun 2015 09:44:51 +0000 (10:44 +0100)]
Fix masked_variance overflow for non-highbitdepth

Change-Id: I668b034111ccc877c70cf65035a9c748fe00aeba

10 years agoMerge "Windows build fix" into nextgen
Debargha Mukherjee [Fri, 5 Jun 2015 17:40:05 +0000 (17:40 +0000)]
Merge "Windows build fix" into nextgen

10 years agoWindows build fix
Debargha Mukherjee [Fri, 5 Jun 2015 17:24:24 +0000 (10:24 -0700)]
Windows build fix

Change-Id: I663e0d61c4115a5297baeb25f39743e1eea95fbf

10 years agoFix lossless mode in tx-skip experiment
hui su [Thu, 4 Jun 2015 21:13:26 +0000 (14:13 -0700)]
Fix lossless mode in tx-skip experiment

Change-Id: I205fa8614a9079ce7e6ca87d7b08fe3602bbd376

10 years agoAdded tests for high bit depth quantize functions
Julia Robson [Thu, 4 Jun 2015 14:48:51 +0000 (15:48 +0100)]
Added tests for high bit depth quantize functions

This was originally part of change-id:Idef18f90b111a0d0c9546543d3347e551908fd78
but the rest of that patch has previously been incorporated into nextgen
without these tests.

Change-Id: I6ac491ed1cfc153e0ed3cb999c76feac9d5e74e3

10 years agoFix lossless mode with new_quant experiment
Debargha Mukherjee [Thu, 4 Jun 2015 09:45:54 +0000 (02:45 -0700)]
Fix lossless mode with new_quant experiment

Change-Id: Ia0b00746e3f7aacf8d9488db3522ab72d9dc22fd

10 years agointrabc: Remove NEWDV from the intra mode tree.
Alex Converse [Wed, 6 May 2015 21:01:10 +0000 (14:01 -0700)]
intrabc: Remove NEWDV from the intra mode tree.

Change-Id: I4e13da4bf70386f6f70ea733fb9adeb41682c735

10 years agointrabc: Fix mismatch introduced by 85a7485.
Alex Converse [Wed, 3 Jun 2015 20:07:49 +0000 (13:07 -0700)]
intrabc: Fix mismatch introduced by 85a7485.

Change-Id: I1ea51ebbd1e3cb60a13a30e3252253493c28baec

10 years agoMerge "intrabc: Include NEWDV in mode_2_counter" into nextgen
Alex Converse [Wed, 3 Jun 2015 16:09:35 +0000 (16:09 +0000)]
Merge "intrabc: Include NEWDV in mode_2_counter" into nextgen

10 years agoMerge "Implied skip is 0 for intra blocks" into nextgen
Debargha Mukherjee [Tue, 2 Jun 2015 21:10:17 +0000 (21:10 +0000)]
Merge "Implied skip is 0 for intra blocks" into nextgen

10 years agoMerge "intrabc: Add odd-pel support for 444." into nextgen
Alex Converse [Tue, 2 Jun 2015 19:03:14 +0000 (19:03 +0000)]
Merge "intrabc: Add odd-pel support for 444." into nextgen

10 years agointrabc: Include NEWDV in mode_2_counter
Alex Converse [Tue, 2 Jun 2015 05:09:01 +0000 (22:09 -0700)]
intrabc: Include NEWDV in mode_2_counter

This fixes a failure to decode as VP9 with frame level signalling
patched in and removed to implictly force it off.

This is primarily a correctness fix, but there were small coding gains.
0.074% smaller on derlf
0.180% smaller on screen_content

Change-Id: I4cc264bf4d9016201924be198a0c5b8374020fd9

10 years agoImplied skip is 0 for intra blocks
Pieter Kapsenberg [Fri, 29 May 2015 00:09:52 +0000 (17:09 -0700)]
Implied skip is 0 for intra blocks

Skip flag is removed from the bitstream for all blocks coded in intra
mode. Very minor coding gain in derf and stdhd sets (0.048 and 0.1)

Change-Id: I79f03300f16d6fa84ce54405cafecab8a021cd7d

10 years agoFix the upper limit of row/column tile index
Jingning Han [Fri, 29 May 2015 16:09:33 +0000 (09:09 -0700)]
Fix the upper limit of row/column tile index

If the decoder is configured to decode a tile indexed above the
upper limit, the internal codec will clip the index to the upper
limits.

Change-Id: Icbc1bb7b14069ac009e0a2042dd66a46d6f76679

10 years agoAllow vpxdec to produce target tile reconstruction
Jingning Han [Fri, 29 May 2015 06:16:35 +0000 (23:16 -0700)]
Allow vpxdec to produce target tile reconstruction

This commit allows the vpxdec to produce the target tile
reconstruction as its output. When the provided tile index is -1,
the entire corresponding row/column tiles will be reconstructed.
If the tile index is over the upper limit, the decoder will decode
left-most/bottom tile.

Change-Id: I4c18bdb32099f736f99b8842f7f177a32b3fee09

10 years agoEnable selective key frame tile decoding
Jingning Han [Wed, 27 May 2015 17:48:26 +0000 (10:48 -0700)]
Enable selective key frame tile decoding

This commit allows the decoder to decode selective tiles according
to decoder configuration settings. To decode a single tile out of
the provided key frame bit-stream (test_kf.webm), set compiler
configuration:

--enable-experimental --enable-row-tile --enable-key-frame-tile

use the command:

vpxdec -o test_dec.y4m test_kf.webm --tile-row=1 --tile-column=2

where the tile's row and column indexes are 1 and 2, respectively.

To decode all row tiles inside a provided column index, use:
--tile-row=-1 --tile-column=2

To decode all column tiles inside a provided row index, use:
--tile-row=2 --tile-column=-1

Change-Id: Ib73c266414dcee7eaab5d741b90d0058970dae56

10 years agointrabc: Add odd-pel support for 444.
Alex Converse [Thu, 28 May 2015 21:35:23 +0000 (14:35 -0700)]
intrabc: Add odd-pel support for 444.

Odd pel still needs more testing on 4:2:0.

0.2% smaller on the PNG set. 4095 better, 1324 worse.

Change-Id: Idf2f108a96462c7696e63d533605e4a926a8b203

10 years agoMerge "Intrabc high bit depth functionality" into nextgen
Debargha Mukherjee [Thu, 28 May 2015 22:40:37 +0000 (22:40 +0000)]
Merge "Intrabc high bit depth functionality" into nextgen

10 years agoMerge "Turn on global rotation-zoom warping prediction" into nextgen
Debargha Mukherjee [Thu, 28 May 2015 22:07:29 +0000 (22:07 +0000)]
Merge "Turn on global rotation-zoom warping prediction" into nextgen

10 years agoIntrabc high bit depth functionality
Julia Robson [Thu, 21 May 2015 16:53:14 +0000 (17:53 +0100)]
Intrabc high bit depth functionality

Fixing compile errors when intrabc and vp9_highbitdepth are both enabled.
Error arose because intrabc uses vp9_setup_scale_factors_for_frame()
which takes an additional argument (use_highbitdepth) when
vp9_highbitdepth is enabled.

Change-Id: I6a15b09dcea0d35525f4c25efb6848804ae5cfab

10 years agoMerge "palette: don't repeat calc_dist for j = 0" into nextgen
Alex Converse [Thu, 28 May 2015 01:55:04 +0000 (01:55 +0000)]
Merge "palette: don't repeat calc_dist for j = 0" into nextgen

10 years agoMerge "Fix a memory overflow issue in large scale tile use case" into nextgen
Jingning Han [Thu, 28 May 2015 01:53:55 +0000 (01:53 +0000)]
Merge "Fix a memory overflow issue in large scale tile use case" into nextgen

10 years agoMerge "Configure selective tile decoding" into nextgen
Jingning Han [Thu, 28 May 2015 01:53:38 +0000 (01:53 +0000)]
Merge "Configure selective tile decoding" into nextgen

10 years agopalette: don't repeat calc_dist for j = 0
Alex Converse [Wed, 27 May 2015 18:27:26 +0000 (11:27 -0700)]
palette: don't repeat calc_dist for j = 0

Change-Id: I043c3b9696dbfa76436763ef871600cb54ade212

10 years agoTurn on global rotation-zoom warping prediction
Debargha Mukherjee [Mon, 18 May 2015 23:23:24 +0000 (16:23 -0700)]
Turn on global rotation-zoom warping prediction

With this patch, the ZEROMV mode is overloaded to represent
a single global dominant motion using one of three models:
1. True zero translation motion (as before)
2. A translation motion different from 0
3. A Rotation-zoom affine model where the predictor is warped
The actual model used is indicated at the frame level for
each reference frame.
A metric that computes the ratio of the error with a global
non-zero model to the error for zero motion, is used to
determine on the encoder side whether to use one of the two
non-zero models or not.

Change-Id: I1f3d235b8860e543191237024a89041ff3aad689

10 years agoFix a memory overflow issue in large scale tile use case
Jingning Han [Wed, 27 May 2015 18:15:32 +0000 (11:15 -0700)]
Fix a memory overflow issue in large scale tile use case

Increase the tok_count array size defined in VP9_COMP to avoid
out of boundary memory access.

Change-Id: Ib277c1102cb5f8355b1d69991532a94e24a2242e

10 years agoMerge "Fixed a bug in palette expt for 444 video format" into nextgen
hui su [Wed, 27 May 2015 18:14:15 +0000 (18:14 +0000)]
Merge "Fixed a bug in palette expt for 444 video format" into nextgen

10 years agoFixed a bug in palette expt for 444 video format
hui su [Wed, 27 May 2015 01:10:05 +0000 (18:10 -0700)]
Fixed a bug in palette expt for 444 video format

The bug was that palette color map was not correctly recorded.

Change-Id: I237494d1d6449c53c5d24eed8c25dcf2f8d21e66

10 years agoMerge "Make the tile coding syntax support large scale tile decoding" into nextgen
Jingning Han [Wed, 27 May 2015 16:40:48 +0000 (16:40 +0000)]
Merge "Make the tile coding syntax support large scale tile decoding" into nextgen

10 years agoConfigure selective tile decoding
Jingning Han [Wed, 27 May 2015 00:09:34 +0000 (17:09 -0700)]
Configure selective tile decoding

Add decoder control to extract single tile from key frame.

Change-Id: Id018bf267d748bfd96a6261055de4194f632190c

10 years agoMerge "intrabc: Fix probs of illegal modes." into nextgen
Alex Converse [Wed, 27 May 2015 01:03:01 +0000 (01:03 +0000)]
Merge "intrabc: Fix probs of illegal modes." into nextgen

10 years agoMake the tile coding syntax support large scale tile decoding
Jingning Han [Tue, 26 May 2015 19:29:49 +0000 (12:29 -0700)]
Make the tile coding syntax support large scale tile decoding

This commit makes the bit-stream syntax support fast selective tile
decoding in a large scale tile array. It reduces the computational
complexity of computing the target tile offset in the bit-stream
from quadratic to linear scale, while maintaining relatively small
stack space requirement (in the order of 1024 bytes instead of 1M
bytes). The overhead cost due to tile separation remains identical.

Change-Id: Id60c6915733d33a627f49e167c57d2534e70aa96

10 years agointrabc: Fix probs of illegal modes.
Alex Converse [Fri, 8 May 2015 18:48:57 +0000 (11:48 -0700)]
intrabc: Fix probs of illegal modes.

ibc: derflr -0.069 screen_content -0.085
ibc+tx+pal: derflr +0.017 screen_content -0.029

Change-Id: I79c0ca0a5c38c53afd209286dbcd78873e070098

10 years agoFix assertion conditions for tile coding
Jingning Han [Tue, 26 May 2015 15:58:34 +0000 (08:58 -0700)]
Fix assertion conditions for tile coding

Allow up to 1024 row/col tiles instead of 4 for row tile and 64 for
column tile.

Change-Id: I1702254e241401821f3aaab1645f6df6dbc48b8d

10 years agoMerge "Loop_postfilter high bit depth functionality" into nextgen
Debargha Mukherjee [Mon, 25 May 2015 08:04:03 +0000 (08:04 +0000)]
Merge "Loop_postfilter high bit depth functionality" into nextgen