CPIPS: Learning to Preserve Perceptual Distances in End-to-End Image Compression
This addresses the need for efficient image compression in IoT, drones, and self-driving cars, where machines process visual data, but it is incremental as it builds on existing neural codecs and perceptual metrics.
The paper tackles the problem of creating compressed image representations that are efficient for both human vision and machine vision tasks, resulting in CPIPS, a method that computes perceptual similarity significantly faster than existing DNN-based metrics.
Lossy image coding standards such as JPEG and MPEG have successfully achieved high compression rates for human consumption of multimedia data. However, with the increasing prevalence of IoT devices, drones, and self-driving cars, machines rather than humans are processing a greater portion of captured visual content. Consequently, it is crucial to pursue an efficient compressed representation that caters not only to human vision but also to image processing and machine vision tasks. Drawing inspiration from the efficient coding hypothesis in biological systems and the modeling of the sensory cortex in neural science, we repurpose the compressed latent representation to prioritize semantic relevance while preserving perceptual distance. Our proposed method, Compressed Perceptual Image Patch Similarity (CPIPS), can be derived at a minimal cost from a learned neural codec and computed significantly faster than DNN-based perceptual metrics such as LPIPS and DISTS.