]>
granicus.if.org Git - libvpx/log
Dmitry Kovalev [Thu, 25 Apr 2013 00:20:53 +0000 (17:20 -0700)]
Handling frame references and scale factors in one for loop.
Using ALLOWED_REFS_PER_FRAME constants instead of hard coded 3, replacing
memcpy with plain struct assignment.
Change-Id: Ibc86f5d175fcb3f3a3eddacf593525370f1f854c
Jingning Han [Tue, 23 Apr 2013 17:12:18 +0000 (10:12 -0700)]
Contextual entropy coding of partition syntax
This commit enables selecting probability models for recursive block
partition information syntax, depending on its above/left partition
information, as well as the current block size. These conditional
probability models are reasonably stationary and consistent across
frames, hence the backward adaptive approach is used to maintain and
update the contextual models.
It achieves coding performance gains (on top of enabling rectangular
block sizes):
derf: 0.242%
yt: 0.391%
hd: 0.376%
stdhd: 0.645%
Change-Id: Ie513d9673337f0d27abd65fb566b711d0844ec2e
Ronald S. Bultje [Wed, 24 Apr 2013 18:45:26 +0000 (11:45 -0700)]
Merge "Change chroma loopfilter to skip inner SB edges for tx16x16 also." into experimental
Ronald S. Bultje [Wed, 24 Apr 2013 18:45:19 +0000 (11:45 -0700)]
Merge "Minor indent changes in loopfilter code." into experimental
Ronald S. Bultje [Wed, 24 Apr 2013 18:45:13 +0000 (11:45 -0700)]
Merge "Add basic building blocks for 8x8 superblocks experiment." into experimental
Ronald S. Bultje [Tue, 23 Apr 2013 21:12:08 +0000 (14:12 -0700)]
Change chroma loopfilter to skip inner SB edges for tx16x16 also.
Change-Id: I6ea9e110b5c5b07ab7d092886dbd51a6eccc0217
Paul Wilkins [Wed, 24 Apr 2013 16:44:48 +0000 (09:44 -0700)]
Merge "Extension of segmentation to 8 segments." into experimental
Paul Wilkins [Wed, 24 Apr 2013 16:43:01 +0000 (09:43 -0700)]
Merge "Simplify Segment Coding" into experimental
Deb Mukherjee [Wed, 24 Apr 2013 15:57:25 +0000 (08:57 -0700)]
Merge "Fix in token allocation with zerogroup expt" into experimental
Paul Wilkins [Wed, 24 Apr 2013 12:04:45 +0000 (13:04 +0100)]
Extension of segmentation to 8 segments.
Also some further simplification following removal
of top node code.
There is an issue in regards to the shared file vp8cx.h
in regard to the roi_map as this interface assumes that
there are only 4 segments. I have left the value here as
4 for now meaning that the roi_map interface is broken
for VP9.
Note that this change would have been easier if I hadn't
had to search for hard wire instances of the number 4
and <= 3.
Change-Id: Ia8b6deea4be4dbd20deb1656e689dd43a5f190e8
Paul Wilkins [Wed, 24 Apr 2013 09:20:52 +0000 (10:20 +0100)]
Simplify Segment Coding
Remove top node optimization.
The improvement this gives is not sufficient to justify
the extra complexity.
Change-Id: I2bb4a12a50ffd52cacfa4a3e8acbb2e522066905
Paul Wilkins [Tue, 23 Apr 2013 13:01:55 +0000 (14:01 +0100)]
Simple implicit segmentation experiment.
Change-Id: Iaef16122732c2a81e0927f9862b51b68dc788712
John Koleszar [Wed, 24 Apr 2013 02:21:30 +0000 (19:21 -0700)]
Merge "Move skip_block from BLOCK to MACROBLOCK" into experimental
John Koleszar [Wed, 24 Apr 2013 02:21:17 +0000 (19:21 -0700)]
Merge "Move quantizer data from BLOCK to MACROBLOCK" into experimental
Dmitry Kovalev [Wed, 24 Apr 2013 01:03:29 +0000 (18:03 -0700)]
Merge "Code cleanup inside vp9_get_pred_context function." into experimental
Dmitry Kovalev [Wed, 24 Apr 2013 01:02:46 +0000 (18:02 -0700)]
Merge "Simple cleanup inside vp9_decodframe.c and vp9_entropymode.c." into experimental
Dmitry Kovalev [Wed, 24 Apr 2013 01:01:37 +0000 (18:01 -0700)]
Merge "Removing redundant code in vp9_entropymode.c." into experimental
John Koleszar [Tue, 23 Apr 2013 22:25:46 +0000 (15:25 -0700)]
Move skip_block from BLOCK to MACROBLOCK
This data is fixed at the MB level, so move it to the common part
of MACROBLOCK.
Change-Id: Idd8c87118e501cdf0a202bd84c28b502a8234edf
John Koleszar [Tue, 23 Apr 2013 22:22:18 +0000 (15:22 -0700)]
Move quantizer data from BLOCK to MACROBLOCK
Quantizers can vary per plane, but not per block. Move these values to
the per-plane part of MACROBLOCK.
Change-Id: I320a55e38b7b28b29aec751a4aca5ccd0c9b9326
John Koleszar [Wed, 24 Apr 2013 00:42:12 +0000 (17:42 -0700)]
Merge "Remove coeff from BLOCK" into experimental
John Koleszar [Wed, 24 Apr 2013 00:41:59 +0000 (17:41 -0700)]
Merge "Convert coeff to per-plane MACROBLOCK data" into experimental
Jingning Han [Wed, 24 Apr 2013 00:16:16 +0000 (17:16 -0700)]
Merge "Enable rectangular support for comp inter-intra" into experimental
John Koleszar [Tue, 23 Apr 2013 21:38:52 +0000 (14:38 -0700)]
Remove coeff from BLOCK
Lookup the data per-plane from the MACROBLOCK struct.
Change-Id: I9253c4d3cf886aa9ab4aeab23a2156bfcf994ede
John Koleszar [Tue, 23 Apr 2013 16:51:09 +0000 (09:51 -0700)]
Convert coeff to per-plane MACROBLOCK data
This commit moves the coeff storage from the MACROBLOCK struct to its
per-plane part. The next commit will remove the coeff member from the
BLOCK structure so that it is consistently accessed per-plane.
Also refactors vp9_sb_block_error_c and vp9_sb_uv_block_error_c to be
variable subsampling aware.
Change-Id: I18c30f87f27c3a012119b6c1970d5fa499804455
John Koleszar [Tue, 23 Apr 2013 23:24:08 +0000 (16:24 -0700)]
Merge "Move src_diff to per-plane MACROBLOCK data" into experimental
Dmitry Kovalev [Tue, 23 Apr 2013 23:18:09 +0000 (16:18 -0700)]
Code cleanup inside vp9_get_pred_context function.
Change-Id: Id06b7a299a26ed944a401faae51907537f722a7e
Dmitry Kovalev [Tue, 23 Apr 2013 22:56:27 +0000 (15:56 -0700)]
Removing redundant code in vp9_entropymode.c.
Change-Id: Ia7266b8d3aa3d5cff2db0c3b2f014def045759af
Dmitry Kovalev [Tue, 23 Apr 2013 22:50:56 +0000 (15:50 -0700)]
Simple cleanup inside vp9_decodframe.c and vp9_entropymode.c.
Change-Id: I62dde981f5201c5fbc22001609ee4b5fd0a9bdf5
Jingning Han [Tue, 23 Apr 2013 22:07:35 +0000 (15:07 -0700)]
Enable rectangular support for comp inter-intra
This commit enables rectangular block prediction of compound
inter-intra mode. It combines the mb/sb32/sb64 prediction functions
into a unified version with configurable block width and height.
This fixes the enc/dec mismatch of the codebase when
comp-interintra-pred is enabled.
Change-Id: I1d0db2f1f184007802df04fcd12b9dadb3189ff0
Ronald S. Bultje [Tue, 23 Apr 2013 22:29:19 +0000 (15:29 -0700)]
Merge changes I54acef34,I72d42971 into experimental
* changes:
Make some sb_type comparisons independent of literal enum values.
Make loopfilter aware of rectangular blocks.
Dmitry Kovalev [Tue, 23 Apr 2013 20:53:14 +0000 (13:53 -0700)]
Merge "Adding select_txfm_size function + vp9_decode_frame cleanup." into experimental
Jingning Han [Tue, 23 Apr 2013 20:52:50 +0000 (13:52 -0700)]
Merge "Remove the unused gf_active_ptr from MACROBLOCK" into experimental
Jingning Han [Tue, 23 Apr 2013 19:59:33 +0000 (12:59 -0700)]
Remove the unused gf_active_ptr from MACROBLOCK
The variable gf_active_ptr is not used anywhere, hence is removed.
Change-Id: I6badc20fade48ae12fd19e7e4cd8fe5e051c20ac
Dmitry Kovalev [Tue, 23 Apr 2013 19:44:51 +0000 (12:44 -0700)]
Merge "Adding get_scan_{4x4, 8x8, 16x16} functions." into experimental
Ronald S. Bultje [Tue, 23 Apr 2013 19:37:05 +0000 (12:37 -0700)]
Minor indent changes in loopfilter code.
Change-Id: I0cdc951558e4d7748f63df8c03b1c9dce086acb0
Ronald S. Bultje [Tue, 23 Apr 2013 19:33:18 +0000 (12:33 -0700)]
Add basic building blocks for 8x8 superblocks experiment.
Change-Id: I274a1d2e461e6ffdb106bac4ad6951692ace314e
Ronald S. Bultje [Tue, 23 Apr 2013 18:29:12 +0000 (11:29 -0700)]
Make some sb_type comparisons independent of literal enum values.
Change-Id: I54acef342b8e787e05af0febd7cf0d7d10288383
Ronald S. Bultje [Tue, 23 Apr 2013 19:32:32 +0000 (12:32 -0700)]
Make loopfilter aware of rectangular blocks.
Also use explicitely named enum values in sb_type comparisons, rather
than relying on absolute integer values, because enum values may
change in the future.
Change-Id: I72d42971a98157af93413a25ac2c7e6f9b369cec
John Koleszar [Tue, 23 Apr 2013 15:26:10 +0000 (08:26 -0700)]
Move src_diff to per-plane MACROBLOCK data
First in a series of commits making certain MACROBLOCK members
addressable per-plane. This commit also refactors the block subtraction
functions vp9_subtract_b, vp9_subtract_sby_c, etc to be
loops-over-planes and variable subsampling aware.
Change-Id: I371d092b914ae0a495dfd852ea1a3d2467be6ec3
Ronald S. Bultje [Tue, 23 Apr 2013 18:19:39 +0000 (11:19 -0700)]
Remove unused corner_predictor and log2_minus_1 functions.
Change-Id: Ic659544ca12b1bc191b93ddfa378964bd133bfc9
Ronald S. Bultje [Tue, 23 Apr 2013 18:18:42 +0000 (11:18 -0700)]
Remove unused stuffing function.
Change-Id: I2bc8d775f8d698bf8582f4eecabc2329452e8d9b
Scott LaVarnway [Tue, 23 Apr 2013 17:22:13 +0000 (10:22 -0700)]
Merge "Eliminated prev_mip memsets/memcpys" into experimental
Deb Mukherjee [Tue, 23 Apr 2013 15:51:31 +0000 (08:51 -0700)]
Fix in token allocation with zerogroup expt
Fix to increase allocation of tokens at very high rates.
Change-Id: Ia27aa0316b0fab664230800f9c9947b5c68ecd58
Deb Mukherjee [Tue, 23 Apr 2013 06:24:06 +0000 (23:24 -0700)]
Removing configure option for implicit wtd pred
Change-Id: Ibe0628905925fc81bbb192651276f417c2aa7552
Deb Mukherjee [Tue, 23 Apr 2013 06:22:28 +0000 (23:22 -0700)]
Merge "Removing the implicit compound inter experiment" into experimental
Deb Mukherjee [Mon, 22 Apr 2013 23:29:34 +0000 (16:29 -0700)]
Removing the implicit compound inter experiment
Removing this experiment for now, since it has been broken with
the latest code changes.
Change-Id: I1be2181b56de490fcb577f5905b5e147a8ed82d8
Jim Bankoski [Mon, 22 Apr 2013 23:42:33 +0000 (16:42 -0700)]
Merge "new version of speed 1" into experimental
Jim Bankoski [Mon, 22 Apr 2013 22:42:41 +0000 (15:42 -0700)]
new version of speed 1
This version of speed 1 only disables modes at higher resolution that
had distortions >2x the best mode we found...
The hope is that this could be a replacement for speed 0 ...
Change-Id: I7421f1016b8958314469da84c4dccddf25390720
Dmitry Kovalev [Mon, 22 Apr 2013 21:39:21 +0000 (14:39 -0700)]
Adding select_txfm_size function + vp9_decode_frame cleanup.
Change-Id: I7594ca84106ee48f83e6fc3098c0de7a28d5a183
Dmitry Kovalev [Mon, 22 Apr 2013 21:08:41 +0000 (14:08 -0700)]
Adding get_scan_{4x4, 8x8, 16x16} functions.
Change-Id: Id4306ef6d65d4a3984aed50b775bdf48d4f6c438
Deb Mukherjee [Mon, 22 Apr 2013 19:59:51 +0000 (12:59 -0700)]
Merge "Bugfix from reordering frame probs patch" into experimental
Deb Mukherjee [Mon, 22 Apr 2013 19:43:12 +0000 (12:43 -0700)]
Bugfix from reordering frame probs patch
This fixes an intermittent mismatch issue cause by moving
the lossless mode decoding bit to after the loop filter
setup information. We need to ensure that the lossless bit
is decoded prior to loop filter setup.
Change-Id: I3faa3fff8e1013b7405dac91268350e059ed121e
John Koleszar [Sat, 20 Apr 2013 21:17:57 +0000 (14:17 -0700)]
Remove vp9_recon_intra_mbuv
Use common vp9_recon_sbuv instead.
Change-Id: I146f79adfdfda2b52257a52fa783727f12afa246
John Koleszar [Sat, 20 Apr 2013 20:50:22 +0000 (13:50 -0700)]
Rewrite vp9_recon_sb*
Rewrite vp9_recon_sb{,y,uv} to be a loop over planes.
Change-Id: Ica2bbbb3105a1d29b2ff2ead07b76cde9683154c
John Koleszar [Sat, 20 Apr 2013 02:16:14 +0000 (19:16 -0700)]
Move pre, second_pre to per-plane MACROBLOCKD data
Continue moving framebuffers to per-plane data.
Change-Id: I237e5a998b364c4ec20316e7249206c0bff8631a
Deb Mukherjee [Mon, 22 Apr 2013 18:53:14 +0000 (11:53 -0700)]
Merge "Removes the code_nonzerocount experiment" into experimental
Deb Mukherjee [Mon, 22 Apr 2013 17:58:49 +0000 (10:58 -0700)]
Removes the code_nonzerocount experiment
This patch does not seem to give any benefits.
Change-Id: I9d2b4091d6af3dfc0875f24db86c01e2de57f8db
Deb Mukherjee [Mon, 22 Apr 2013 17:33:12 +0000 (10:33 -0700)]
Merge "End of orientation zero group experiment" into experimental
Deb Mukherjee [Thu, 28 Mar 2013 17:42:23 +0000 (10:42 -0700)]
End of orientation zero group experiment
Adds an experiment that codes an end-of-orientation symbol
for every eligible zero encountered in scan order.
This cleans out various other sub-experiments that were part
of the origiinal patch, which will be later included if found
useful.
Results are slightly positive on all sets (0.1 - 0.2% range).
Change-Id: I57765c605fefc7fb9d1b57f1b356843602abefaf
John Koleszar [Fri, 19 Apr 2013 17:45:50 +0000 (10:45 -0700)]
reconinter: remove unnecessary functions, params
Removes the redundant dst pointers from vp9_build_inter_predictors_sb{y,uv}
and the remaining mb specific functions.
Change-Id: I7b6bf439d9394b85ea79b4fe61a3ffc1025720da
Paul Wilkins [Mon, 22 Apr 2013 12:10:36 +0000 (05:10 -0700)]
Merge "make DC_PRED for i4x4 to use real pixels only" into experimental
Sami Pietilä [Mon, 22 Apr 2013 10:13:34 +0000 (03:13 -0700)]
Merge "Reordering frame header probs." into experimental
Dmitry Kovalev [Sat, 20 Apr 2013 18:05:50 +0000 (11:05 -0700)]
Merge "Removing get_segment_id function and using existing vp9_get_pred_mb_segid." into experimental
Dmitry Kovalev [Sat, 20 Apr 2013 18:04:06 +0000 (11:04 -0700)]
Merge "Finally removing BOOL_DECODER and using vp9_reader instead." into experimental
Dmitry Kovalev [Sat, 20 Apr 2013 18:03:24 +0000 (11:03 -0700)]
Merge "Renaming vp9_extra_bit_struct to vp9_extra_bit." into experimental
John Koleszar [Fri, 19 Apr 2013 23:33:45 +0000 (16:33 -0700)]
Merge "Move dst to per-plane MACROBLOCKD data" into experimental
John Koleszar [Fri, 19 Apr 2013 23:33:09 +0000 (16:33 -0700)]
Merge "Remove vp9_recon_mb{,y}" into experimental
Yaowu Xu [Fri, 19 Apr 2013 16:33:07 +0000 (09:33 -0700)]
make DC_PRED for i4x4 to use real pixels only
Wherever there are real pixels available before falling back to use
assumed values 127 and 129.
This also make DC_PRED for i4x4 consistent with DC_PRED for larger
blocks.
Change-Id: I54372924826118da023f402c802ac6ce0caa70c3
John Koleszar [Fri, 19 Apr 2013 23:17:42 +0000 (16:17 -0700)]
Merge "Remove redundant pointers from void vp9_recon_sb{y,uv}" into experimental
John Koleszar [Fri, 19 Apr 2013 22:52:17 +0000 (15:52 -0700)]
Move dst to per-plane MACROBLOCKD data
First in a series of commits moving the framebuffers pointers to
per-plane data, so that they can be indexed numerically rather than
by name.
Change-Id: I6e0d60fd4d51e6375c384eb7321776564df21775
Scott LaVarnway [Fri, 19 Apr 2013 20:32:15 +0000 (16:32 -0400)]
Eliminated prev_mip memsets/memcpys
For 1080 material, this buffer is currently 2,270,928 bytes. This patch swaps
ptrs instead of copying and uses the last show_frame flag instead of setting
the entire buffer to zero. For the test clip used, the decoder improved by up
to 1%.
Change-Id: I686825712ad56043e09ada9808dc489f875a6ce0
Paul Wilkins [Fri, 19 Apr 2013 21:31:19 +0000 (14:31 -0700)]
Merge "Clean out some legacy code." into experimental
Dmitry Kovalev [Fri, 19 Apr 2013 21:25:32 +0000 (14:25 -0700)]
Removing get_segment_id function and using existing vp9_get_pred_mb_segid.
Change-Id: Iff35d4b2f8f65511f80c594958c01fb4673fa033
Paul Wilkins [Fri, 19 Apr 2013 15:56:14 +0000 (16:56 +0100)]
Clean out some legacy code.
Removed some unused legacy code relating to GF activity.
Change-Id: I0b0fe90a1a5208aaa81d7393ba6cf75ab412dbda
Yunqing Wang [Fri, 19 Apr 2013 21:12:43 +0000 (14:12 -0700)]
Merge "Remove unused parameters in handle_inter_mode" into experimental
Paul Wilkins [Fri, 19 Apr 2013 21:09:44 +0000 (14:09 -0700)]
Merge "Mv ref candidates cut to 2." into experimental
John Koleszar [Fri, 19 Apr 2013 19:12:00 +0000 (12:12 -0700)]
Remove vp9_recon_mb{,y}
Use the common sb functions instead.
Change-Id: I4fa0a8ee3c6ada56271dd09bf895b97642f55858
John Koleszar [Fri, 19 Apr 2013 19:06:07 +0000 (12:06 -0700)]
Remove redundant pointers from void vp9_recon_sb{y,uv}
Remove the unnecessary _s_ from their names, and add a new
vp9_recon_sb() that calls the y and uv variants.
Change-Id: I7ffaa5ff5605a8472cac2a53de8cf889353039a6
John Koleszar [Fri, 19 Apr 2013 19:04:23 +0000 (12:04 -0700)]
Merge "Move diff to MACROBLOCKD per-plane data." into experimental
Dmitry Kovalev [Fri, 19 Apr 2013 17:44:03 +0000 (10:44 -0700)]
Segmentation cleanup, adding {set, get}_segment_id functions.
Change-Id: I55c2688e06ae5d7dfccc1b1983f233ab1c7978db
Dmitry Kovalev [Fri, 19 Apr 2013 18:14:33 +0000 (11:14 -0700)]
Renaming vp9_extra_bit_struct to vp9_extra_bit.
Change-Id: Ie4713da125e954c1d30e1d4cbeb38666fce90ccc
John Koleszar [Fri, 19 Apr 2013 18:07:47 +0000 (11:07 -0700)]
Move diff to MACROBLOCKD per-plane data.
Change-Id: Ic27af09e38af8317ac4743241883d577a44f1490
Dmitry Kovalev [Fri, 19 Apr 2013 17:37:24 +0000 (10:37 -0700)]
Finally removing BOOL_DECODER and using vp9_reader instead.
Change-Id: I03d5b6f22f0930893709c6db5f1b06762ad3354e
John Koleszar [Wed, 17 Apr 2013 20:41:18 +0000 (13:41 -0700)]
make build_inter_predictors block size agnostic (split)
All build_inter_predictors can now be serviced by the same inner
function.
Change-Id: I40b08bee8f047286db4b1aad9dcae37b879c3f2a
John Koleszar [Wed, 17 Apr 2013 19:24:51 +0000 (12:24 -0700)]
Removing rounding from UV MV calculation for SPLITMV
Similar to the prior change that removed the rounding from non-SPLITMV
modes. Improves quality by a similar amount (Additional +0.087% on derf)
Change-Id: I39d80b4a3037a3aa7e285eb2320346ddaf646f52
John Koleszar [Fri, 19 Apr 2013 17:23:04 +0000 (10:23 -0700)]
Merge "make buid_inter_predictors block size agnostic (chroma)" into experimental
John Koleszar [Fri, 19 Apr 2013 17:22:54 +0000 (10:22 -0700)]
Merge "Use SSSE3 for 2d filters larger than 16" into experimental
Dmitry Kovalev [Fri, 19 Apr 2013 17:09:53 +0000 (10:09 -0700)]
Merge "Replacing BOOL_DECODER with vp9_reader inside vp9_decodframe." into experimental
Dmitry Kovalev [Fri, 19 Apr 2013 17:09:04 +0000 (10:09 -0700)]
Merge "Fixing member names inside TOKENVALUE and TOKENEXTRA structs." into experimental
Jim Bankoski [Fri, 19 Apr 2013 16:57:38 +0000 (09:57 -0700)]
Merge "catch all for new block sizes" into experimental
John Koleszar [Fri, 19 Apr 2013 15:24:33 +0000 (08:24 -0700)]
Use SSSE3 for 2d filters larger than 16
The C code was being used as a fallback for the >16 case, but only for 2D.
Change-Id: I1e2e6da9e4b28bd88bde9ba4dd32724ce466cf6f
Jim Bankoski [Fri, 19 Apr 2013 16:48:56 +0000 (09:48 -0700)]
catch all for new block sizes
Just make sure we don't stop them from testing in speed 1.
Change-Id: Iec9b3dba0a32616ff7a451207e0f54b81bb72575
Jim Bankoski [Fri, 19 Apr 2013 15:28:35 +0000 (08:28 -0700)]
Merge "set up a new speed 1" into experimental
Paul Wilkins [Fri, 19 Apr 2013 14:40:36 +0000 (15:40 +0100)]
Mv ref candidates cut to 2.
Further simplification of mvref search to return
only the top two candidates. Distance weights removed
as the test order reflects distance anyway.
Change-Id: I0518cab7280258fec2058670add4f853fab7b855
Jim Bankoski [Fri, 19 Apr 2013 15:04:57 +0000 (08:04 -0700)]
set up a new speed 1
slightly worse results for faster encodes
Change-Id: I25ea82a18ce20635dbcd328808c1d05ac1f58fd7
Paul Wilkins [Fri, 19 Apr 2013 10:54:02 +0000 (11:54 +0100)]
Removal of CONFIG_NEW_MVREF experiment.
This experiment has failed to give much benefit but
does add complexity so deprecated.
Change-Id: Ic7b929ba706390b9907ef0b4f965bd401ca799a4
Paul Wilkins [Fri, 19 Apr 2013 09:25:36 +0000 (10:25 +0100)]
Remove dummy place holder function.
void __attribute__((noinline)) hi(void) { }
Causes build failure in VS2008
Change-Id: Ie2f2a09d90bd5502c492e4d9f4983532a0edbc01
Paul Wilkins [Fri, 19 Apr 2013 10:10:16 +0000 (11:10 +0100)]
Simplification of MVref search.
As we are no longer able to sort the candidate
mvrefs in both encoder and decode and given
that the cost of explicit signalling has proved
prohibitive, it no longer makes sense to find more
than 2 candidates.
This patch:
Modifies and simplifies add_candidate_mv()
Removes the forced addition of a 0 vector in the
MAX_MV_REF_CANDIDATES-1 position (in preparation
to reducing MAX_MV_REF_CANDIDATES to 2).
Re-orders the addition of candidates slightly.
This actually gives small gains (circa 0.2% on std-hd)
A subsequent patch will remove NEW_MVREF experiment,
reduce MAX_MV_REF_CANDIDATES to 2 and remove distance
weights as these are implicit now in the order.
Change-Id: I3dbe1a6f8a1a18b3c108257069c22a1141a207a4
Paul Wilkins [Fri, 19 Apr 2013 09:27:22 +0000 (02:27 -0700)]
Merge "Adjustments to key frame sizing." into experimental
Paul Wilkins [Tue, 16 Apr 2013 17:59:39 +0000 (18:59 +0100)]
Adjustments to key frame sizing.
Adjustments take heavier account of the frame near a kf
in deciding boost and limit the total number that can contribute.
Also adjusted the minq calculations such that in most cases we
generate a smaller key frame.
Modified the code that accounts for how static the sequence is and
added some adjustment based on image size. This is still very
crude but smaller images tend to behave better with a larger
delta between KF Q and other frames than larger image formats.
Changes give sizable gains in overall PSNR on all the test sets but the
biggest gains (~3%) were on the std-hd set.
The gains were smaller for SSIM but still significant.
Average PSNR results are mixed because this metric can very easily
be altered by having a very good / lossless coding of one or two frames.
Some of the YT and YT-HD clips in particular have blank lead ins and
allowing lossless coding of these appears to make a big difference to
average PSNR but it reality does not help much at all.
Change-Id: I6bfe485a1d330b47c783832f1717c95c535464ec