Improving AVIF in Open Source

10 months ago

AVIF is gaining popularity for web images due to its compression and quality, but open-source encoders faced consistency and efficiency issues.
Gianni Rosato, founder of Halide Compression, worked on improving AVIF through the SVT-AV1-PSY project, focusing on still image optimization.
Key improvements in AVIF encoding include quantization matrix scaling, deblocking loop filter sharpness, variance-adaptive quantization, and screen content detection.
The 'Tune Still Picture' mode in SVT-AV1-PSY introduced techniques that improved compression by up to 15% and enhanced perceptual quality.
Modern image codecs on the Web are often derived from video standards, leading to suboptimal image encoding performance.
Halide Compression aims to create an image-first ecosystem with efficient licensable encoders and expert consulting to address current limitations.
Future developments like AV2 could further enhance still image performance, but challenges remain in encoder consistency and resource allocation.

Hasty Briefsbeta