← All resources

Lossless Float Image Compression

Auto-discovered

Blog Intermediate

Aras Pranckevičius

Visit resource ↗

Description

Lossless Float Image Compression is a benchmarking blog post by Aras Pranckevičius on aras-p.info. The URL is one article comparing lossless codecs on multi-layer floating-point EXR-style images used in film and VFX workflows.

It tests OpenEXR ZIP and upcoming HTJ2K, JPEG-XL lossless via libjxl at several effort levels, and a custom chunked scheme using meshoptimizer vertex coding optionally followed by zstd. The data set includes Blender multi-channel renders and Poly Haven HDRIs with mixed half and float channels. Charts compare ratio versus compress and decompress throughput on Apple Silicon and Ryzen Windows, with notes on threading and FP16 round-trip limits in libjxl.

Readers should be intermediate engineers who store float image buffers and care about size and speed. Familiarity with EXR or HDR pipelines helps.

The article is free, with harness source linked from the post. Measurements are in-memory encode and decode without disk I/O.

It helps choose a lossless float image path and see where JPEG-XL and HTJ2K sit relative to ZIP and a meshoptimizer-based packer.