ANFIC: Image Compression Using Augmented Normalizing Flows

This paper introduces an end-to-end learned image compression system, termed ANFIC, based on Augmented Normalizing Flows (ANF). ANF is a new type of flow model, which stacks multiple variational autoencoders (VAE) for greater model expressiveness. The VAE-based has gone mainstream, showing promising performance. Our work presents the first attempt to leverage in flow-based framework. ANFIC advances further efficiency by stacking and extending hierarchically VAE’s. invertibility ANF, together with our training strategies, enables support wide range quality levels without changing encoding decoding networks. Extensive experimental results show that terms PSNR-RGB, performs comparably or better than state-of-the-art compression. Moreover, it close VVC intra coding, from low-rate up perceptually lossless In particular, achieves performance, when extended conditional convolution variable rate single model. source code can be found at https://github.com/dororojames/ANFIC .