This commit aligns the rate-distortion metrics for both luma and
chroma components in super transform rate-distortion optimization.
It improves the coding gains due to var-tx and supertx experiments
by 0.2% for high resolution test sets.
Change-Id: Ib89d99e29cb5ee27b1f867e301954d4164d8b364
This commit aligns the rate-distortion metric for the recursive
transform block partitioning and the super transform. It resolves
the conflicts between these two experiments. The coding performance
gains of the combined experiments (var-tx + super-tx) has been
improved:
derf 0.89% -> 1.9%
hevcmr 1.06% -> 1.8%
stdhd 0.29% -> 1.4%
hevchr 0.80% -> 2.3%
Change-Id: I7e33994ad70c1b2751435620815f867d82172f41
This patch makes rd optimization use the same context for computing
the rate cost of coding the partitioning as the packer actually uses
when emitting it in write_modes_sb.
Change-Id: Idb1427bb2f9c37ab80c6aa182f7ff754ef0595cb
This commit enables entropy coding for dynamic reference motion
vector modes. The probability model is contexted on the ranking
categories of the reference motion vector candidates.
Change-Id: I09b58d98a409d63ec1a407331e29f8945b7ef17d
BD-rate performance improvement (on top of ext-intra):
derflr 0.22%
hevclr 0.36%
hevcmr 0.48%
hevchr 0.37%
stdhd 0.19%
Average speed impact on some derf clips is about 40% slower (on
top of ext-intra). Speed improvment is a to-do.
Change-Id: I8fe3fe8c5e4f60d0462778adbcc15c84dfbe7a25
Update blk_skip in update_state_supertx() and rd_supertx_sb().
Performance (SUPERTX+VAR_TX): TBD
(Eventually will merge update_state() and update_state_supertx())
Change-Id: I34ef982b80151ba2dfba745859cb2ca7b90dc888
Fixes integer pel MV usage for the sub8x8 case, which fixes a
rare mismatch issue.
Also adds some other minor missing code related to filter threshes.
Change-Id: I6b07e6cf9b287ba4b5bd6599af4a7412e50b3bdc
There is still an assertion failure when tokenizing transform
coefficients for a supertx block, due to the eob not being set
consistently with the coefficients, so we always recode supertx blocks
for now. Also added further PICK_MODE_CONTEXT instances to avoid
potential clash between horizontal/vertical/split partition SUPERTX
trials.
Change-Id: I5f3da1fa0d8d20fc21face170487e1a285fd1cc6
The encoder did not update left_txfm_context and above_txfm_context in
MACROBLOCKD (used for choosing the probability context for the vartx
split bits) when the supertx bit was set for a block. The deoder on the
other hand did update these for supertx blocks. The encoder used these
to compute the context counts, which the packer then uses to adapt it's
probabilities. This results in the packer and the decoder using
different probabilities.
This patch harmonizes the encoder and the decoder by making the encoder
update the mentioned context for supertx coded blocks.
Change-Id: I3a22132124b1bce2ee501d640ceab374b19e3ca1
This is necessary when using SUPERTX, as the bitstream packer relies on
tx_size being set correctly to decide whether to output the block using
supertx or not.
Change-Id: I79e776b3b810f4a15b9dbc6afdd6fc90c73c8934
The loop filter relies on inter_tx_size in MB_MODE_INFO being set
properly when VAR_TX is enabled. Supertx coded blocks did not set this
previously at all, and the differing garbage values eventually resulted
in in a YUV mismatch between encoder and decoder after loop filtering.
This patch fixes this by setting inter_tx_size to the proper supertx
size in both the encoder and the decoder. This should also mean that
loop filtering is done at the proper transform boundaries, even when
supertx or vartx is being used.
Change-Id: I41a564cd6d34ce4a8313ad4efa89d905f5ead731
Fixes some of the issues introduced by a merge from master.
derflr: -0.893% BDRATE
hevcmr: -1.667% BDRATE
Change-Id: I4c5b40ec63a6f19521191d3c730af87db3c4bc00
Combinations of different mv modes for two reference frames
are allowed in compound inter modes. 9 options are enabled,
including NEAREST_NEARESTMV, NEAREST_NEARMV, NEAR_NEARESTMV,
NEAREST_NEWMV, NEW_NEARESTMV, NEAR_NEWMV, NEW_NEARMV, ZERO_ZEROMV,
and NEW_NEWMV.
This experiment is mostly deported from the nextgen branch.
It is made compatible with other experiments
Coding gain of EXT_INTER(derflr/hevcmr/hevchd): 0.533%/0.728%/0.639%
Change-Id: Id47e97284e6481b186870afbad33204b7a33dbb0
NEW2MV is enabled, representing a new motion vector predicted from
NEARMV. It is mostly ported from nextgen, where it was named
NEW_INTER.
A few fixes are done for sub8x8 RDO to correct some misused
mv references in the original patch.
A 'bug-fix' for encoding complexity is done, reducing the additional
encoding time from 50% to 20%. In sub8x8 case, the old patch
did motion search for every interpolation filter (vp9 only
searches once). This fix also slightly improves the coding gain.
This experiment has been made compatible with REF_MV and EXT_REFS.
Coding gain (derflr/hevcmr/hevchd): 0.267%/0.542%/0.257%
Change-Id: I9a94c5f292e7454492a877f65072e8aedba087d4
The nominal tx_type for a given mode is used as a context
to encode the actual tx_type for intra.
Results:
derflr: -0.241% BDRATE
hevcmr: -0.366% BDRATE
Change-Id: Icfe7b0a58d79bc6497a06e3441779afec6e01e21
It allows the codec to account for certain corner cases when
processing inter prediction mode entropy coding.
Change-Id: Ied451f4fff26ba579f6556554b8381ff2ccd0003
Various additional changes were made to make the experiment
compatible with misc_fixes.
derflr: +0.979%
hevcmr: +0.865%
Speed-wise with --enable-supertx the encoder is only about 10%
slower than without. Decoding impact is about 30% slowdown.
Note this does not work with ext-tx or var-tx yet. That is
a TODO.
Change-Id: If25af4241a7a9efbd28f58eda3c4f044c7a7ef4b
This commit allows the codec to analyze the reference motion vector
candidate list and adaptively reduce the size of inter prediction
mode set.
Change-Id: Ied6a403843b860d66f26ed485c1825c05c71bdfc
This commit re-works the entropy coding scheme of the motion
compensated prediction modes. It allows more flexible hyperplane
partition for precise classification.
Change-Id: Iba5035c76691946cf1386b6c495e399c3d9c8fc5
This change has been imported from VP9 and
alters the nature and use of exhaustive motion search.
Firstly any exhaustive search is preceded by a normal step search.
The exhaustive search is only carried out if the distortion resulting
from the step search is above a threshold value.
Secondly the simple +/- 64 exhaustive search is replaced by a
multi stage mesh based search where each stage has a range
and step/interval size. Subsequent stages use the best position from
the previous stage as the center of the search but use a reduced range
and interval size.
For example:
stage 1: Range +/- 64 interval 4
stage 2: Range +/- 32 interval 2
stage 3: Range +/- 15 interval 1
This process, especially when it follows on from a normal step
search, has shown itself to be almost as effective as a full range
exhaustive search with step 1 but greatly lowers the computational
complexity such that it can be used in some cases for speeds 0-2.
This patch also removes a double exhaustive search for sub 8x8 blocks
which also contained a bug (the two searches used different distortion
metrics).
For best quality in my test animation sequence this patch has almost
no impact on quality but improves encode speed by more than 5X.
Restricted use in good quality speeds 0-2 yields significant quality gains
on the animation test of 0.2 - 0.5 db with only a small impact on encode
speed. On most natural video clips, however, where the step search
is performing well, the quality gain and speed impact are small.
Change-Id: Iac24152ae239f42a246f39ee5f00fe62d193cb98
Under the experiment of EXT_REFS: LAST2_FRAME, LAST3_FRAME, and
LAST4_FRAME.
Coding efficiency: derflr +1.601%; hevchr +1.895%
Speed: Encoder slowed down by ~75%
Change-Id: Ifeee5f049c2c1f7cb29bc897622ef88897082ecf
Remove MISC_FIXES flags except for the changes on MV precision, which
has a 0.1% performance drop.
On derflr, the impact is -0.012%.
Change-Id: I0a74e5a212dd0cb827192a318c92a714c9681e45
Adds a new interpolation experiment.
Improves entropy coding to send the filter type only if
the motion vectors have subpel components.
Adds one new 8-tap smooth filter, and tweaks the others.
derflr: +0.695%
hevcmr: +0.305%
About 5% encode slowdown. No visible impact for decoding.
Also makes the interpolation framework flexible to support both
strictly interpolating filters as well as non-interpolating
filters that filter integer offsets. This is mainly for
further experimentation and if not found useful the code will
be removed.
Change-Id: I8db9cde56ca916be771fe54a130d608bf10786e6
Currently there are two parts in this experiment: extra directional intra
prediction modes and the filter intra modes migrated from the nextgen branch.
Several macros are defined in "blockd.h" to provide controls of the experiment
settings. Setting "DR_ONLY" as 1 (default is 0) means we only use directional
modes, and skip the filter-intra modes; "EXT_INTRA_ANGLES" (default is 128)
defines the number of different angles we want to support; setting
"ANGLE_FAST_SEARCH" as 1 (default is 1) means we use fast sub-optimal search
for the best prediction angle, instead of exhaustive search. The fast search
is about 6 times faster than the exhaustive search, while preserving about
60% of the coding gains.
With extra directional prediction modes (fast search), we observe the following
code gains (number in parentheses is for all-key-frame setting):
derflr +0.42% (+1.79%)
hevclr +0.78% (+2.19%)
hevcmr +1.20% (+3.49%)
stdhd +0.56%
Speed-wise, about 110% slower for key frames, and 30% slower overall.
The gains of filter intra modes mostly add up with the gains of directional
modes. The overall coding gain of this experiment:
derflr +0.94%
hevclr +1.46%
hevcmr +1.94%
stdhd +1.58%
Change-Id: Ida9ad00cdb33aff422d06eb42b4f4e5f25df8a2a
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
