GitLab

(1) Work with var-refs to remove redundant bits in ref frame
    coding;
(2) Add a new uni-directional compound reference pair:
    (LAST_FRAME, LAST3_FRAME);
(3) Redesign the contexts for encoding uni-directional reference frame
    pairs;
(4) Use aom_entropy_optimizer to collect stats for all the default
    probability setups related to the coding of reference frames.

Compared against the baseline (default enabled tools excluding ext-tx
and global-motion for encoder speed concern) with one-sided-compound,
the coding gain of ext-comp-refs + var-refs - one-sided-compound is:

lowres: avg_psnr -0.385%; ovr_psnr -0.378% ssim -0.344%
midres: avg_psnr -0.466%; ovr_psnr -0.447% ssim -0.513%

AWCY - High Latency:
   PSNR | PSNR Cb | PSNR Cr | PSNR HVS |    SSIM | MS SSIM | CIEDE 2000
-0.2758 | -0.1526 | -0.0965 |  -0.2581 | -0.2492 | -0.2534 |    -0.2118

AWCY - Low Latency:
   PSNR | PSNR Cb | PSNR Cr | PSNR HVS |    SSIM | MS SSIM | CIEDE 2000
-1.0467 | -1.4500 | -0.9732 |  -0.9928 | -1.0407 | -1.0180 |    -1.0049

Compared against the baseline (default enabled tools excluding ext-tx
and global-motion for encoder speed concern) without
one-sided-compound, the coding gain of
ext-comp-refs + var-refs - one-sided-compound is:

lowres: avg_psnr -0.875%; ovr_psnr -0.877% ssim -0.895%
midres: avg_psnr -0.824%; ovr_psnr -0.802% ssim -0.843%

Change-Id: I8de774c9a74c20632ea93ccb0c17779fa94431cb

This is intended to be a no-op when scaling is not
enabled, but is expected to result in more accurate
prediction when references need to be scaled.

However note all xs, yx, subpel_x and subpel_y values
are now at higher than 1/16th precision.

Change-Id: I4b22573ea290a31fc58ead980bb0d5e5a9e89243

Prioritize modes that are more likely to be picked. Impact on
compression is neutral. Keyframe encoding speed increases by about 5%.

Change-Id: I57f7481a42b833203e4cfb84806fe755ec37d615

Change-Id: Id5c3df752528a3ea74e655b3d18a5d6f258a2238

Define the syntax and entropy coding templates for
NCOBMC_ADAPT_WEIGHT. The actual values of the default
probabilities and the index tree structure need to
be fine tuned.

In this experiment all mv's in a superblock are sent
first as in the ncobmc case.

Change-Id: I68d50d3d27346c2847ea449a1168c6a99fbb4d3d

cb4x4 itself should not require these sizes.

This simplifies compatibility with other experiments, since we can
first make them work with cb4x4 (which is now on by default), and
then worry about chroma_sub8x8 and chroma_2x2 (which is not) in
separate steps.

Encoder and decoder output should remain unchanged.

Change-Id: Iff2a5494cab3b7d96f881e8bd9cd4bf18c817cfa

The tool of ext-comp-refs adds the uni-directional compound reference
prediction. In details, 3 pairs of uni-direcitonal compound references
are added for the comp ref prediction:
(LAST_FRAME, LAST2_FRAME),
(LAST_FRAME, GOLDEN_FRAME), and
(BWDREF_FRAME, ALTREF_FRAME).

This new tool of ext-comp-refs will eventually overwrite
one-side-compound and have the two coding tools to merge to one.

It achieves -0.35 ~ -0.55% coding gains in BDRate, compared against
AV1 baseline with the default experiments on, but without
one-sided-compound. It achieves -0.2% ~ -0.3% coding gains when
one-sided-compound is on. It achieves larger gains on higher
resolution.

Change-Id: Icbdb16e97b96aaebaf2213f5f72d5331e2e358eb

RESIZE_SCALE_DENOMINATOR and SUPERRES_SCALE_DENOMINATOR were two
constants with the same value that did essentially the same thing.

This patch merges the two into SCALE_DENOMINATOR for simplicity's sake.

Change-Id: I252a9b7f89f10d77bdb0c3cf2d67d31d337afa4b

Unifies the command line arguments for resize and superes.
Removes resize arguments that are currently unused.

Change-Id: I1362dffceb8257cee539f92679d1e8d79df08e66

Adds "superres-mode" and "superres-numerator" arguments to configure
superres scaling.

There are only two modes right now: SUPERRES_FIXED and SUPERRES_DYNAMIC.
0 sets the fixed mode and 1 sets the dynamic mode.

For the fixed mode superres will scale to the provided numerator where
possible. The cumulative scale change over resizing and superres is
limited to 1/2, so it may have to back off.

For the dynamic mode, the provided numerator will be used as the initial
numerator. Asides from that it's just random for now.

Change-Id: Ibae01c77abba2599fbf51096459d09ddca458d4f

Only apply quantisation matrices when the transform is 2D.

Change-Id: Iae9ac910c11199f7944a80d10d334db3b711059d

Now the single ref comp mode should work with WEDGE and
COMPOUND_SEGMENT. For motion_var, the OBMC_CAUSAL mode uses the 2nd
predictor if the neighboring block is single ref comp mode predicted.

This patch removes the mode of SR_NEAREST_NEWMV and leaves four
single ref comp modes in total:

SR_NEAREST_NEARMV
SR_NEAR_NEWMV
SR_ZERO_NEWMV
SR_NEW_NEWMV

Change-Id: If6140455771f0f1a3b947766eccf82f23cc6b67a

Earlier, the condition was negating all non-directional modes to check
if a mode is directional. This was error-prone, e.g. when a new
non-directional mode is added.

By checking for directional modes directly, we avoid such errors.

Change-Id: Ia4a62e278cd73078c53ed5096db646eff77f054e

cb4x4 itself should not require these sizes.

This simplifies compatibility with other experiments, since we can
first make them work with cb4x4 (which is now on by default), and
then worry about chroma_2x2 (which is not) in separate steps.

Encoder and decoder output should remain unchanged.

Change-Id: I4e9fcdae49f238b5099a3c74a398fe993c2545f8

Separate the codes into a table of distinct values
and an index into that table.
Pull the SSE calculation of the RDO loop and avoid
repeating for the same alpha values.

Change-Id: I8c4bd7eab6f8000e6aca9687d9190abc3e270c37
Signed-off-by: David Michael Barr <b@rr-dav.id.au>

Currently the "transform units" used to measure transform sizes
in some parts of the code are based on the smallest defined
transform size.

cb4x4 currently defines a 2x2 transform size, even when chroma_2x2
is not enabled, which means that the scale of the transform units
was always double that of MODEINFO units.

Several areas of the code were hard-coding this assumption instead
of converting from one to the other using appropriate constants.

Change-Id: Ibc55671aa5bc3ad272cb8a036f9c4f9621df85ab

Create named constants for the minimum TX_SIZE used in the luma
plane, as well as the minimum allowed TX_SIZE for which we have to
explicitly code a selection (i.e., one larger than the minimum luma
TX_SIZE).

Then, use these constants consistently with tx_size_to_depth() to
index the tx_size counts in FRAME_COUNTS, and also consistently use
TX_SIZE named constants to index the tx_size_implied counts.

Failing to index these counts correctly can, e.g., cause tx_mode to
be chosen incorrectly.

Change-Id: I706a62a33e2282e67c97a68bade87fb8023ec13b

It gives 0.1% gain on lowres and midres

Change-Id: I555a492a68571c525713840d73aa5614fe80a87d

Improved the performance a little bit compared to using the full set

Change-Id: I28a96459fa77d7e1d5c101d162faa2effd525341

Change-Id: I640cf08ac583ff5092a18e81cd0e33f05609db6b

Use CONFIG_INTERINTRA to properly separate interintra from the basic
ext_inter experiment.
When macro REDUCE_INTERINTRA_MODES is 1, only 4 ii modes are enabled
so as to reduce the complexity overhead.
(Right now the flag is off)

Change-Id: Iec6f36a1000f181adbb822ad66c7d5b2625e9cc7

This has been taken care of already.

Change-Id: I0198fdbcd48fc355013de79644513279c8acb0de

This experiment extends ALT_INTRA by adding two new modes:
smooth horizontal and smooth vertical.

Improvement on *intra frames* in BDRate (PSNR):
===============================================

AWCY (high latency): -0.46%
(Also, -1.0% or more on PSNR Cb,Cr and APSNR Cb,Cr).

AWCY (low latency): -0.43%
(Also, -0.88% to -0.94% on PSNR Cb,Cr and APSNR Cb,Cr).

Google sets:
lowres: -0.454
midres: -0.484
hdres:  -0.525

Improvement on *video overall* in BDRate (PSNR):
================================================

AWCY (high latency): -0.15%

Google sets:
lowres: -0.085
midres: -0.079

Change-Id: I9f4e7c1b8ded1fe244c72838f336103ccc715d50

Change-Id: Id490c57df5dca13c59a62b0a542779784842944c

The angle estimation in ext-intra didn't account for
alt-intra.

Also use c99 variable declaration style.

BUG=aomedia:444

Change-Id: I001b0278eb8cf560975c873a423de024c10d08b4

Writes and reads alpha to and from the bitstream.

A special case is needed on the encoder side to handle prediction block
skips. Since whether or not a prediction block is skipped during CfL, a
rollback is required if the block was skipped and the alpha index was
not zero. The advantage of this is that no signaling is required when
the prediction block is skipped as it is assumed tha the alpha index is
zero.

A encode facade is added to the intra prediction facade as CfL requires
special encoder side operations.

Change-Id: Ic3b11d0fdbd51389d862112eb09d8785127a6b06

Change-Id: I96e386678f6227f3ca735c9be2725cb9795b3296

The frame superresolution experiment currently makes no use of the
features of the loop restoration experiment. While this is planned for
the future, there is no reason to continue having frame superresolution
depend on loop restoration for now.

Frame superresolution was made dependent by having every preproccesor if
containing its config flag also depend on the loop restoration config
flag. Those if's are modified to only check the frame superresolution
flag by this patch.

Change-Id: Ieef47de29125af8d080739ffeac5c55a1fa99e23

The hookup with the encoder-decoder pipeline will be in the following
CLs. This is under the experiment of compound-singleref, with ext-
inter enabled.

Change-Id: I6523a1a6525539eb7c6aa6c5e11724a694e8bf6c

To avoid confusion, inspection now returns INTRA_INVALID for as the
uv_mode when the block type is inter.

Change-Id: I48c64586078bb1ef1ca256a9f3c5c2bfb4a2abc3

Change-Id: I8caa3e0ca51b733b3af87d24ee511f6bb52222d1

Change-Id: I9b6f99768c60e743d2fadfaabc532e7c15dc40f3

This reverts commit 5ca25dfc.

It breaks high-bitdepth builds with:
--enable-aom-highbitdepth --enable-experimental --enable-dual-filter
--enable-entropy --enable-ext-inter --enable-ext-intra
--enable-ext-partition-types --enable-ext-refs --enable-ext-tx
--enable-filter-intra --enable-loop-restoration

or more simply:
--enable-aom-highbitdepth --enable-experimental --enable-ext-inter

BUG=aomedia:463

Change-Id: If814131e634e4411569859f7426efafe158b83fd

This is the second patch.
* Add delta loop filter level at super block level.
* Add symbol coding for delta lf syntax
* Allow delta lf to work with segments

Change-Id: I393a021a875d03c6e113127cbb3543fc077308e4

Change-Id: I359d100548ed337d643a421591243565b08945ed

Added 2 tile encoding modes:
TILE_NORMAL mode supports the large-scale tile coding up to 1024 x 1024
tiles;
TILE_VR mode supports the large-scale tile coding and also the decoding
of a single tile.

TILE_NORMAL mode allows the non-vr content to be encoded in multiple tiles
without using the coding tool that are designed for vr applications.

Change-Id: Id804806a1bbcb916b7f5dc7e2c5509d88f3defcb

Change-Id: I246da885154303d139f5d97c53ebfc55d6dc82ca

A bit and a three bit literal have been added to the uncompressed frame
header for use by the frame superres experiment.

The bit is true if scaling is to be used, and is immediately followed by
a three bit literal that encodes the scale to use. If the first bit is
false, scaling is disabled and the scale factor numerator is set to the
denominator (ie. 1:1 scaling). No literal follows if scaling is
disabled.

The denominator has been defined as a constant 16. The literal is biased
by a defined constant of 8 - allowing fractions from 1/2 to 15/16
scaling in steps of 1/16 when scaling is used. Experimentation will be
needed to discover which of these are useful.

The bit and literal are immediately after the optional render_width and
render_height, so that the superres parameters can be written and read
just after the regular width and height, without interfering with the
render_size parameters.

This patch also adds an arbitrary write to make the scale 1:1, so as to
not trigger any scaling until it's ready.

Accompanying encode and decode helper functions are added.

Change-Id: I8caa6247c73f5c7f84ef1fde1e80eb9b20bde0e3

Adds the superres scale numerator, width, and height to the AV1_COMMON
struct for later use in encoding and decoding. This value will also be
added to the uncompressed frame header.

A #define statement is also added to represent the denominator - 16.

Change-Id: I4acec39ef3c67fa9f7118dac28daec3a2d734ed5

* Dering and clpf were merged into a single pass.
* 32x32 and 128x128 filter block sizes for clpf were removed.
* RDO for dering and clpf merged and improved:
  - "0" no longer required to be in the strength selection
  - Dering strength can now be 0, 1 or 2 bits per block

              LL    HL
PSNR:       -0.04 -0.01
PSNR HVS:   -0.27 -0.18
SSIM:       -0.15 +0.01
CIEDE 2000: -0.11 -0.03
APSNR:      -0.03 -0.00
MS SSIM:    -0.18 -0.11

Change-Id: I9f002a16ad218eab6007f90f1f176232443495f0