Apple has introduced PICO, an AI-powered image compression codec designed to deliver significantly higher efficiency while maintaining perceptual quality. The technology uses neural networks trained on human vision data to optimize encoding for how people actually see images.
AI compression maintains visual quality while reducing file size
The codec is designed to integrate with current image processing workflows. The technology is optimized for scenarios requiring efficient data handling, including mobile imaging and online media sharing.
The codec achieves a 2.3x to 3x reduction in bitrate compared to industry standards like AV1 and JPEG AI when evaluated under identical conditions. This efficiency gain comes without sacrificing the visual fidelity that users expect from modern imaging systems.
Testing on Apple's iPhone 17 Pro Max demonstrated encoding speeds of 230 milliseconds for a 12-megapixel image, with decoding completing in just 150 milliseconds. These performance metrics position PICO as faster than most learning-based codecs currently running on NVIDIA V100 hardware.
The evaluation methodology combined objective metrics including MS-SSIM and LPIPS with large-scale human subjective testing to ensure the compression results aligned with perceptual quality goals. The system prioritizes natural detail retention over raw numerical scores, reflecting Apple's focus on real-world viewing experiences.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.