JPEG XS targets low-latency transport of uncompressed video
The JPEG XS standard (ISO/IEC 21122) specifies a compression technology designed for visually lossless compression with low latency and hardware implementation complexity. It supports various pixel formats and offers compression ratios up to 10:1, making it suitable for applications that previously transported uncompressed image data, such as professional video links and IP transport. The specification is multi-part, covering core coding, profiles, transport, conformance testing, and reference software.
Key Takeaways
- JPEG XS supports end-to-end latency of a few lines and hardware implementations that do not require external memory.
- Typical compression ratios are up to 10:1 for 4:4:4, 4:2:2, and 4:2:0 images, with higher ratios possible depending on the image and application.
- The codec supports raw Bayer, RGB, and YUV formats, all up to 16 bits per component, plus lossless compression up to 12 bits per component.
- JPEG XS is specified in five parts: core coding, profiles and buffer models, transport and container, conformance testing, and reference software.
- The standard is explicitly aimed at uses that previously transported uncompressed image data, including 3G/6G/12G-SDI, SMPTE 2022-5/6, and SMPTE 2110 IP transport.
Why It Matters
JPEG XS is built for workflows where latency and bandwidth matter more than maximum compression: it aims for visually lossless output, fraction-of-a-frame delay, and hardware implementations with low complexity. That combination matters for live production chains, IP transport, and other systems that currently carry uncompressed video. The multi-part spec also matters because it separates coding, transport, conformance, and reference software, which should make implementation and interoperability more concrete. Watch for which parts of the standard vendors adopt first: core coding, transport/container support, or the reference software on Windows, MacOS, and Linux.
Read full article at jpeg.org