A Universal Optimization Framework for Learning-based Image Codec

Recently, machine learning-based image compression attracts increasing interests and is approaching the state-of-the-art ratio. But unlike traditional codec, it lacks a universal optimization method to seek efficient representation for different images. In this paper, we develop plug-and-play framework seeking higher ratio, which can be flexibly applied existing potential future networks. To make latent more efficient, propose novel algorithm adaptively remove redundancy each image. Additionally, inspired by of side information codecs, introduce into our framework, integrate with further enhance particular, joint optimization, achieve fine rate control using only single model instead training models rate-distortion trade-offs, significantly reduces storage cost support multiple bit rates. Experimental results demonstrate that proposed remarkably boost achieving than \(10\% \) additional saving on three representative network structures. With \(7.6\% against latest coding standard VVC Kodak dataset, yielding