A new JPEG coding library, Jpegli, has been introduced by Google to enhance image compression and quality on the internet.
Jpegli maintains strong backward compatibility with the original JPEG standard while offering improved capabilities and a 35% compression ratio enhancement at high-quality compression settings.
The library employs adaptive quantization heuristics, refined quantization matrix selection, precise intermediate results, and the option to utilize a more advanced color space to reduce noise and enhance image quality.
These techniques have been meticulously designed to utilize the traditional 8-bit JPEG framework, ensuring compatibility with existing JPEG viewers.
Jpegli also supports dynamics of 10+ bits per component, accessible through an API extension, which reduces visible banding artifacts in slow gradients.
Jpegli’s coding speed is comparable to traditional methods, allowing for seamless integration into existing workflows without compromising performance or memory usage.
Crowdsourcing raters were engaged to compare image quality among Jpegli, libjpeg-turbo, and MozJPEG at various bitrates, with Jpegli receiving a higher ELO rating than libjpeg-turbo at a 32% higher bitrate.
Jpegli has the potential to accelerate the internet and enhance visual appeal by compressing high-quality images more efficiently than traditional JPEG codecs.
Source: https://opensource.googleblog.com/2024/04/introducing-jpegli-new-jpeg-coding-library.html
