Introduce --max-tx-size
Description
--max-tx-size allows the encoder to restrict selection of blocks transform sizes up to a maximum size. Valid values are 32 and 64.
In the AV1 standard, 64-pt transforms have the last 32 highest-frequency coefficients zeroed out during encoding, which means coded blocks can look visually blurry, especially when encoding fine noise-like textures.
PSNR and SSIM-based RDO metrics don't seem to detect this blurriness, so this setting combats this issue by not allowing 64-pt transforms to be considered in the first place. The result is an overall increase in output quality consistency, especially for still images in the medium to high quality range.
Example:
| --max-tx-size 64 | --max-tx-size 32 |
|---|---|
 |
 |
Notice how the background camera noise for --max-tx-size 32 is consistently better well preserved compared to --max-tx-size 64.
Issue
Part of the SVT-AV1-PSY mainlining effort (#2269)
Author(s)
Performance impact
-
quality -
memory -
speed -
8 bit -
10 bit -
N/A
Test set
-
obj-1-fast can be found here -
other -
N/A
Merge method
-
Allow the maintainer to squash and merge when PR is ready to create a 1-commit to the master branch. The maintainer will be able to fix typos / combine commit messages to create a more readable 1-commit message or use whatever is stated in the 'Description' section -
I will clean up my commits and the maintainer shall use 'rebase and merge' to the master branch