GitLab

Change-Id: I9d3f0380865790d8adcb85f03305b193bc1949d7

Only noise level changes to PSNR U/V and bitrate observed, but this is
much more logical/consistent.

Change-Id: I31be92bbbb416b57f3a23f3da6ed2c9ac6a29955

Change-Id: Ie4382b8a1c0f87ce50e9afefd1cef8ca55435c61

Various fixes for pvq build.

Change-Id: Ideebdb072ed5786f3224e93ded5ec75a23e68dab

Reduce the context model size for key frame modes from 30240 bits
to 4500 bits, i.e., less than 1/6 of the original context model.
The coding performance loss on key frame is 0.14% for lowres and
noise level difference for video sequence. The loss on key frame
for midres is 0.05% and noise level for whole video. The change
on hdres kf coding is 0.015%.

Change-Id: I9e36825e5c5ee6ba35038c3ca349ad1ad3429910

Change-Id: Ic843e99bd9b79cb9a0a26b95e3a48717ff2ec2a5

Change-Id: Ifa983d83a509cdfad78f6400df7d60c8f5b4f68c

Change-Id: Ieb28f40d85e4db4af33648c32c406dd2931ceb89

With ext-partition and tx64x64 enabled, the encoder could choose
TX_64X64 to encode the transform for a subsampled plane of a
BLOCK_128X128 block. This broke an assumption in the nested loop in
write_tokens_b and also caused bug 827 (with a rather cryptic
failure).

This patch changes get_vartx_max_txsize to ensure that the encoder and
decoder don't think they can use TX_64X64 in this situation. It also
adds a couple of assertions to the loop mentioned above so that if
something comes unstuck it'll be much more obvious what went wrong.

BUG=aomedia:827

Change-Id: Ie093f2f20f6242949d68e950c8f95b100867ee17

This function is given a bsize and an mi array and has to figure out
what partition to use to divide the given bsize in the direction of
the sizes it finds in the mi array. (Since each block size can be
reached by only one sequence of partitions, this can be done
unambiguously)

The previous version was correct, working by looking up entries in the
partition_lookup array. Unfortunately, that lookup isn't quite enough
when CONFIG_EXT_PARTITION_TYPES is true, so it then had to do some
slightly confusing computations to fix things up after the fact.

The new version should be more self-explanatory and doesn't work by
looking things up in a magic array. It looks up the width and height
corresponding to bsize and compares them with the width and height
corresponding to the sb_type at mi_row,mi_col in the mi array. When
CONFIG_EXT_PARTITION_TYPES is false, this is all you need, and the
four corresponding cases can be found by a lookup in an array of 4
elements.

With extended partition types and a sufficiently large block, you have
to do a bit more searching. For example, if bsize is BLOCK_16X16 and
the subsize is BLOCK_8X8, the partition might be PARTITION_SPLIT, but
it might be one of PARTITION_HORZ_A or PARTITION_VERT_A instead. The
new code adds some comments to explain what's going on.

A nice side-effect of rewriting get_partition in this way is that it
lets us completely dispense with the partition_lookup array.

The patch also fixes comments for the A/B extended partitions in
enums.h, which were slightly backwards (a "horizontal split means two
blocks vertically above one another)

Change-Id: I4b48189103aa63e1859f25a15d7690d53ca7baf5

When updating default_partition_cdf, this sums the probabilities that
were divided evenly across the pairs PARTITION_HORZ_A/PARTITION_HORZ_B
and PARTITION_VERT_A/PARTITION_VERT_B. Those summed probabilities now
get distributed evenly across the triples you get by adding
PARTITION_HORZ_4 and PARTITION_VERT_4, respectively.

Rather than implement 2X8/8X2 blocks for now, ss_size_lookup returns
4X8/8X4 block sizes to use as chroma transform sizes for 4X16/16X4
blocks.

The changes in setup_pred_plane and set_skip_context are because this
is presumably the first time we've had to deal with 16x4 or 4x16
blocks. Since BLOCK_16X4 is not less than BLOCK_8X8, the existing
logic didn't work (and the "shuffle back one" logic should probably be
done for small widths and heights separately).

Change-Id: If28d8954da42d6c726f2bcce2cb5242154b0870c

Also add another default probabilities and for two-modes

Change-Id: Ia503301e6d8e89c8d7c945ff25588807107d4807

Change-Id: I0c3772110e9fa62ac687bd99e290b5006bf3bd6c

Change-Id: Ie2c34490dc50cb242bcd701308e6b55243883b15

This patch adds support for 4:1 rectangular blocks to various common
data arrays, and adds new partition types to the EXT_PARTITION_TYPES
experiment which will use them.

This patch has the following restrictions, which can be lifted in
future patches:

  * ext-partition-types is incompatible with fp_mb_stats and supertx
    for the moment

  * Currently only 32x32 superblocks can use the new partition types

There's a slightly odd restriction about when we allow
PARTITION_HORZ_4 or PARTITION_VERT_4. Since these both live in the
EXT_PARTITION_TYPES CDF, read_partition() can only return them if both
has_rows and has_cols is true. This means that at least half of the
width and height of the block must be visible. It might be nice to
relax that restriction but that would imply a change to how we encode
partition types, which seems already to be in a state of flux, so
maybe it's better to wait until that has settled down.

Change-Id: Id7fc3fd0f762f35f63b3d3e3bf4e07c245c7b4fa

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

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

Make the txfm_above and txfm_left be processed in the unit of
miniumum transform block size. Scale the transform block step
size with respect to the mode_info step size.

Change-Id: Iee4421e005db742cd4ff7899215560063e5f68e5

It gives 0.1% gain on lowres and midres

Change-Id: I555a492a68571c525713840d73aa5614fe80a87d

The hbd transform configurations were originally written for all possible
2d transforms. Now that there are many more possible 2d transforms
due to EXT_TX and RECT_TX, it is simpler to write the cfg for the
4 1D transform types and compose them to make all new possible transform
types. This will allow for an easier integration of the identity transform
for EXT_TX and rectangular transforms for RECT_TX into the current
hbd transform codepath and facilitate the removal of obsolete transforms.
This has no impact on performance.

BUG=aomedia:524

Change-Id: I1e217bcd217fd637b1df94fae62d9c59a0523c1a

Change-Id: I0c58fb5ee2a77ae15f6243b1c645dbe601171e67

BUG=aomedia:396

Change-Id: I1b7f6dddb06eac9e341518a95a0d71465e613e9e

This draft version only pass compiling check, it's not working yet.

The following goal is to use new coding system when doing bitstream
packing but keep old coding system in RD loop.

Change-Id: I224a1581d1cc5c67d73e71558fb77d9faf9c2470

With this patch, --enable-var-tx only enables recursive transform
partitioning without using rectangular transforms.
To enable use of rectangular transforms in addition, use:
--enable-var-tx --enable-rect-tx

The RD selection process is not fully tested under the var-tx flag
only. We might expect certain performance loss there.

Change-Id: Ie6aa17f1bbc3e8563b9990bc9ff79cc860d9a361

Fix an encoding failure issue when var-tx is enabled, while ext-tx
and rect-tx are disabled. This doesn't change coding statistics
when all are enabled.

Change-Id: I4b32387a0a1497380980f8087832aaf6467cdcbe

This commit makes ext-tx and rect-tx experiments supported in the
cb4x4 mode. It resolves an enc/dec mismatch issue when all the
transform experiments are enabled.

The coding gains are
        ext-tx + rect-tx   cb4x4    vartx     total
lowres      4.0%           2.3%      0.5%     6.9%

The encoding speed is about the same when cb4x4 and vartx are
further enabled.

BUG=aomedia:139

Change-Id: I3fdabc6d5de23ceb78ac0751a9bf7332ebc0a3ac

Account for the additional block sizes in these tables.

Change-Id: Iae940f28671714caaf32432940752958ef66f6d5

These are under EXT_TX + RECT_TX experiment combo.

Results
=======

Derf Set:
--------
All Intra frames: 1.8% avg improvement (and 1.78% BD-rate improvement)
Video: 0.230% avg improvement (and 0.262% BD-rate improvement)

Objective-1-fast set
--------------------
Video: 0.52 PSNR improvement

Change-Id: I1893465929858e38419f327752dc61c19b96b997

Also fixes a bug with rectangular transforms

Change-Id: Id459c18d8fdc767678452e0b20c4168a412f4de7

Change-Id: I93493abe3c412fc10f5bb5a2eb157c8db277f4e0

Change the table content to support 4x4 mode_info unit. Refactor
partition_plane_context() to reflect the fact that the minimum
partition block size is 8x8.

Change-Id: Id26fcfba8d74efdb9ae316c9d066cd65358dd7e6

Separate the use cases of number of 8x8 blocks from those of
mode_info array. This allows the mode_info array processing to be
scaled to 4x4 resolution.

Change-Id: Iab78f2540355ce7658d9ea21e902a86f71148d8f

Scale the counting scheme to operate in the unit of 2x2 transform
block size and 4x4 coding block size.

Change-Id: I7c4155c3b29df109278187d5ce0aa78d85a6245e

If the coding block size is 4x4, map the uv transform block size
to 2x2 in non-444 format.

Change-Id: I5767df7c6c1f73938f97745d604207caad6f7d68

clang format gets confused with this table due to #if statements.
Turn off format check on this table.

Change-Id: I21e9069581749d6e01501d01ba779f56207ca84a

Add 2x2 block size syntax to the codec system. It prepares for the
4x4 coding block unit for 420 format. This change retains the same
coding statistics.

Change-Id: If8e9a31bd6b4b75bc994539dc4dd8021d455ba57

Replace the computation of block size in pixels with lookup table
directly.

Change-Id: I39589b2bc1d20372969ff5ad0f60639e64a19b41

Change-Id: Idc44fa29c70c9bb63c9b04cfba679fd05d34fdff

Includes:
Various table updates and fixes to support 64x64 transforms.
Entropymode updates to support tx_size expansion to 64x64.
tx_mode changes to support an ALLOW_64sXx64 transform mode.

Change-Id: Ib9098cfe27d0c015fe3be6ae13e7d09576771b9e