This patch allows the prediction residues of tx-skipped blocks
to use probs that are different from regular transfrom
coefficients for token entropy coding. Prediction residues are
assumed as in band 6.
The initial value of probs is obtained with stats from limited
tests. The statistic model for constrained token nodes has not
been optimized. The probs for token extra bits have not been
optimized. These can be future work.
Certain coding improvment is observed:
derflr with all experiments: +6.26% (+0.10%)
screen_content with palette: +22.48% (+1.28%)
Change-Id: I1c0d78178ee9f3655febb6f30cdaef8ee9f8e3cc
This experiment, referred as NEWMVREF, also merged with NEWMVREF_SUB8X8
and the latter one has been removed. Runborgs results show that:
(1) Turning on this experiment only, compared against the base:
derflf: Average PSNR 0.40%; Overall PSNR 0.40%; SSIM 0.35%
(2) Turning on all the experiments including this feature, compared against
that without this feature, on the highbitdepth case using 12-bit:
derflf: Average PSNR 0.33%; Overall PSNR 0.32%; SSIM 0.30%.
Now for highbitdepth using 12-bit, compared against base:
derflf: Average PSNR 11.12%; Overall PSNR 11.07%; SSIM 20.27%.
Change-Id: Ie61dbfd5a19b8652920d2c602201a25a018a87a6
The basic idea is to use a pixel’s neighboring colors as
context to predict its own color. Up to 4 neighbors are
considered here: left, left-above, above, right-above.
To reduce the number of contexts, the combination of any
4 (or less) colors are mapped to a reduced number of
patterns. For example, 1111, 2222, 3333, … , can be mapped
to the same pattern: AAAA. SImilarly, 1122, 1133, 2233, …,
can be mapped to the pattern AABB. In this way, the total
number of color contexts is reduced to 16.
This almost doubles the gain of palette coding on screen
content videos.
on screen_content
--enable-palette +14.2%
--enable-palette --enable-tx-skip +21.2%
on derflr
--enable-palette +0.12%
with all other experiments +6.16%
Change-Id: I560306dae216f2ac11a9214968c2ad2319fa1718
Also make changes to transmit palette-enabled flag using
neighbor blocks as context.
on screen_content
--enable-palette +7.35%
on derflr
with all other experiments +6.05%
Change-Id: Id6c2f726d21913d54a3f86ecfea474a4044c27f6
on screen_content
--enable-palette +6.74%
on derflr
with all other experiments +6.02%
(--enable-supertx --enable-copy-mode
--enable-ext-tx --enable-filterintra
--enable-tx64x64 --enable-tx-skip
--enable-interintra --enable-wedge-partition
--enable-compound-modes --enable-new-quant
--enable-palette)
Change-Id: Ib85049b4c3fcf52bf95efbc9d6aecf53d53ca1a3
This framework allows lower quantization bins to be shrunk down or
expanded to match closer the source distribution (assuming a generalized
gaussian-like central peaky model for the coefficients) in an
entropy-constrained sense. Specifically, the width of the bins 0-4 are
modified as a factor of the nominal quantization step size and from 5
onwards all bins become the same as the nominal quantization step size.
Further, different bin width profiles as well as reconstruction values
can be used based on the coefficient band as well as the quantization step
size divided into 5 ranges.
A small gain currently on derflr of about 0.16% is observed with the
same paraemters for all q values.
Optimizing the parameters based on qstep value is left as a TODO for now.
Results on derflr with all expts on is +6.08% (up from 5.88%).
Experiments are in progress to tune the parameters for different
coefficient bands and quantization step ranges.
Change-Id: I88429d8cb0777021bfbb689ef69b764eafb3a1de
For 444 videos, a single palette of 3-d colors is
generated for YUV. For 420 videos, there may be two
palettes, one for Y, and the other for UV.
Also fixed a bug when palette and tx-skip are both on.
on derflr
--enable-palette +0.00%
with all experiments +5.87% (was +5.93%)
on screen_content
--enable-palette +6.00%
--enable-palette --enable-tx_skip +15.3%
on screen_content 444 version
--enable-palette +6.76%
--enable-palette --enable-tx_skip +19.5%
Change-Id: I7287090aecc90eebcd4335d132a8c2c3895dfdd4
All stats look fine.
derflr: +0.912 with respect to 10-bit internal baseline
(Was +0.747% w.r.t. 8 bit)
+5.545 with respect to 8-bit baseline
Change-Id: I3c14fd17718a640ea2f6bd39534e0b5cbe04fb66
For all sub8x8 partition, the mv ref has changed to its own nearest_mv
instead of the nearest_mv of the super 8x8 block:
--enable-newmvref-sub8x8: ~0.1% gain for derflr
Besides the above new experiment, code has been cleaned largely for the
sub8x8 motion search, such that the mv ref can be fairly easily changed to
use other options, e.g., the next step's global motion effort.
Change-Id: I8e3f4aaa8553ba8c445369692e079db5ce282593
Added palette coding option for Y channel only, key frame only.
Palette colors are obtained with k-means clustering algorithm.
Run-length coding is used to compress the color indices.
On screen_content:
--enable-palette + 5.75%
--enable-palette --enable-tx_skip +15.04% (was 13.3%)
On derflr:
--enable-palette - 0.03%
with all the other expriments + 5.95% (was 5.98%)
Change-Id: I6d1cf45c889be764d14083170fdf14a424bd31b5
Remove the duplicated step counting the overhead for ext_tx, because
it has been counted in super_block_yrd().
Change-Id: I50bc01d8166572bc0847305565cf33c14ee516e6
Also fixes some build issues with certain experimental combinations.
Results on derflr with all experiments on: +5.516%
Change-Id: I9b492f3d3556bd1f057005571dc9bee63167dd95
COMPOUND_MODES experiment encodes separate MV modes for each frame in a compound
reference prediction. Added modes NEAREST_NEARESTMV, ZERO_ZEROMV,
NEW_NEWMV, NEAREST_NEARMV, NEAR_NEARESTMV, NEW_NEARESTMV, NEAR_NEWMV,
NEW_NEARMV, and NEAREST_NEWMV.
Also enhances the wedge-partition expt to work better with compound
modes.
Results:
derflr +0.227
All experiments on: derflr +5.218
Change-Id: I719e8a34826bf1f1fe3988dac5733a845a89ef2b
A smooth weighting scheme is used to put more weight
on the intra predictor samples near the left/top boundaries
and decaying it to favor the inter predictor samples more as
we move away from these boundaries in the direction of
prediction.
Results:
derflr: +0.609% with only this experiment
derflr: +3.901% with all experiments
Change-Id: Ic9dbe599ad6162fb05900059cbd6fc88b203a09c
In lossless coding, tx_size can be larger than 4x4 when transform
skipping is activated. Compared to regular vp9 lossless coding,
performance improvement for derf is about 5%; gain is larger
for screen content videos.
Change-Id: Ib20ece7e117f29fb91543612757302a2400110b4
This patch improves the non-transform coding mode. At this
point, the coding gain on screen content videos is about
12% for lossless, an 15% for lossy case.
1. Encode tx_skip flags with context. Y tx_skip flag context is
whether the prediction mode is inter or intra. UV flag context
is Y flag.
2. Transform skipping is less helpful when the Q-index is high.
So it is enabled only when the Q-index is smaller than a
threshold. Currently the threshold is set as 255 for intra blocks,
and 0 for inter blocks.
3. The shift of the prediction residue, when copying them to the
coeff buffer, is set as 3 when the Q-index is larger than a
threshold (currently set as 0), and 2 otherwise.
Change-Id: I372973c7518cf385f6e542b22d0f803016e693b0
Reimplements the supertx experiment from the playground branch.
Makes it work with other experiments.
Results:
With --enable-superttx
derflr: +0.958
With --enable-supertx --enable-ext-tx
derflr: +2.25%
With --enable-supertx --enable-ext-tx --enable-filterintra
derflr: +2.73%
Change-Id: I5012418ef2556bf2758146d90c4e2fb8a14610c7
Extends the ext-tx experiment to include regular and flipped
DST variants. A total of 9 transforms are thus possible for
each inter block with transform size <= 16x16.
In this patch currently only the four ADST_ADST variants
(flipped or non-flipped in both dimensions) are enabled
for inter blocks.
The gain with the ext-tx experiment grows to +1.12 on derflr.
Further experiments are underway.
Change-Id: Ia2ed19a334face6135b064748f727fdc9db278ec
Non-transform option is enabled in both intra and inter modes.
In lossless case, the average coding gain on screen content
clips is 11.3% in my test.
Change-Id: I2e8de515fb39e74c61bb86ce0f682d5f79e15188
