[Paper Review] Committee Draft of JPEG XL Image Coding System

Alexander Rhatushnyak, Jan Wassenberg+16 more|arXiv (Cornell University)|Aug 12, 2019

Advanced Data Compression Techniques10 citations

TL;DR

This paper presents JPEG XL, a next-generation image coding system designed for efficient, scalable web delivery and high-fidelity compression. It employs an enhanced block-transform architecture with advanced prediction, transform, and entropy coding to achieve 60% smaller file sizes than JPEG at equivalent visual quality, while supporting lossless, progressive, animated, and reversible transcoding with royalty-free, open-source implementation available in Q4 2019.

ABSTRACT

JPEG XL is a practical approach focused on scalable web distribution and efficient compression of high-quality images. It provides various benefits compared to existing image formats: 60% size reduction at equivalent subjective quality; fast, parallelizable decoding and encoding configurations; features such as progressive, lossless, animation, and reversible transcoding of existing JPEG with 22% size reduction; support for high-quality applications including wide gamut, higher resolution/bit depth/dynamic range, and visually lossless coding. The JPEG XL architecture is traditional block-transform coding with upgrades to each component.

Motivation & Objective

• To design a modern image coding system optimized for web delivery and high-quality imaging needs.
• To reduce file size significantly compared to existing formats like JPEG while maintaining or improving visual quality.
• To support advanced features such as lossless compression, progressive decoding, animation, and reversible transcoding of existing JPEG images.
• To enable efficient encoding and decoding with parallelizable, fast processing for real-time applications.
• To support high dynamic range, wide color gamut, and high bit-depth imaging for professional and future-proof use cases.

Proposed method

• Adopts a traditional block-transform coding architecture with significant enhancements to each component.
• Introduces advanced predictive coding techniques for improved prediction accuracy in spatial and frequency domains.
• Employs a flexible transform scheme with multiple transform types and optimized quantization for perceptual efficiency.
• Uses a sophisticated entropy coding module based on adaptive arithmetic coding with context modeling for high compression efficiency.
• Supports reversible transcoding by enabling lossless conversion from existing JPEG bitstreams with 22% size reduction.
• Designs the system for parallelization in both encoding and decoding to ensure high performance on modern hardware.

Experimental results

Research questions

• RQ1 Can a new image coding standard achieve significantly better compression efficiency than JPEG while maintaining or improving visual quality?
• RQ2 How can a single format efficiently support lossy, lossless, progressive, and animated image coding?
• RQ3 What architectural enhancements to block-transform coding are required to achieve 60% size reduction at equivalent subjective quality?
• RQ4 Can reversible transcoding from existing JPEG be efficiently supported within a new standard?
• RQ5 How can the system be designed for fast, parallelizable encoding and decoding suitable for web-scale deployment?

Key findings

• JPEG XL achieves a 60% reduction in file size compared to JPEG at equivalent subjective image quality, demonstrating superior compression efficiency.
• The system supports reversible transcoding of existing JPEG images, reducing their size by 22% while preserving lossless quality.
• The architecture enables fast, parallelizable encoding and decoding, suitable for real-time and web-based applications.
• The format natively supports high dynamic range, wide color gamut, and high bit-depth imaging for professional and future-proof use.
• A royalty-free, open-source reference implementation was made available in Q4 2019, ensuring broad accessibility and adoption.
• The system supports progressive decoding, animation, and lossless compression as first-class features within a unified framework.