Extended End-to-End optimized Image Compression Method based on a Context-Adaptive Entropy Model

Jooyoung Lee, Seunghyun Cho, Seyoon Jeong, Hyoungjin Kwon, Hyunsuk Ko, Hui Yong Kim, Jin Soo Choi

Abstract: In this paper, we propose an extended compression method using a context-adaptive entropy model. Based on the Lee et al.Ju Hu approach, we extend the network structure so that compression and quality enhancement methods are jointly optimized. In terms of contexts for estimating distributions, we additionally use offset information. By exploiting the extended structure and the additional con-texts, we obtain substantially improved compression performance, in terms of multi-scale structural similarity (MS-SSIM) index, compared to the model without the extensions.