‹ back to 2026-07-22

Wavefront Parallelization for Efficient Learned Image Compression

Research Efficiency & Systems

Merged summary

TL;DR - A training-free wavefront scheduling method accelerates inference in pretrained autoregressive learned image compression models by over 13× without changing rate-distortion performance.

Reorders inference into an optimal staggered wavefront while preserving exact autoregressive dependencies.
Requires no architecture changes or retraining, unlike checkerboard context methods.
Works with existing pretrained models, including Cheng et al.
Offers additional decoding speed by optionally relaxing precise context dependencies.

Sources (1)

Wavefront Parallelization for Efficient Learned Image Compression

arXiv eess.IV Shimon Murai, Fangzheng Lin, Kasidis Arunruangsirilert, Jiro Katto 2026-07-21 arXiv:2607.19082

TL;DR - A training-free wavefront scheduling method accelerates inference in pretrained autoregressive learned image compression models by over 13× without changing rate-distortion performance.

Reorders inference into an optimal staggered wavefront while preserving exact autoregressive dependencies.
Requires no architecture changes or retraining, unlike checkerboard context methods.
Works with existing pretrained models, including Cheng et al.
Offers additional decoding speed by optionally relaxing precise context dependencies.

item →