AVIF images should use bilinear or better chroma upscaling
Categories
(Core :: Graphics: ImageLib, defect, P3)
Tracking
()
People
(Reporter: TD-Linux, Unassigned)
References
(Blocks 1 open bug)
Details
Attachments
(4 files)
AVIF decoding currently uses libyuv for conversion to RGB, which uses box sampling. However, other decoders (including JPEG) use bilinear or better chroma upsampling.
The edges of the red circle should appear blurrier than the original PNG, but not boxy and pixelated. See the version decoded by avifdec, and the JPEG, for examples.
This applies to video using the basic compositor or software libyuv paths as well, but it's not quite as critical as the defects are harder to see in video.
Some early adopters have already noticed this deficiency: https://news.ycombinator.com/item?id=23614807
Reporter | ||
Comment 1•4 years ago
|
||
Reporter | ||
Comment 2•4 years ago
|
||
Reporter | ||
Comment 3•4 years ago
|
||
Updated•4 years ago
|
Comment 4•4 years ago
|
||
There's a similar issue on Chrome: https://bugs.chromium.org/p/chromium/issues/detail?id=1121579
I've filed an enhancement request with libyuv: https://bugs.chromium.org/p/libyuv/issues/detail?id=872
Comment 5•3 years ago
|
||
ITU-T Series H Supplement 18 recommends Lanczos 2 or better. https://hdl.handle.net/11.1002%2F1000%2F13441
Updated•3 years ago
|
Description
•