Deep Residual Learning for Image Compression

TL;DR

This paper introduces a deep residual learning approach combined with sub-pixel convolution for improved learned image compression, demonstrating high MS-SSIM scores at low bitrates.

Contribution

It proposes novel deep residual and sub-pixel convolution techniques specifically tailored for image compression, advancing the state-of-the-art in learned compression methods.

Findings

Achieved 0.972 MS-SSIM at 0.15bpp rate

Developed three models: Kattolab, Kattolabv2, KattolabSSIM

Moderate complexity during validation

Abstract

In this paper, we provide a detailed description on our approach designed for CVPR 2019 Workshop and Challenge on Learned Image Compression (CLIC). Our approach mainly consists of two proposals, i.e. deep residual learning for image compression and sub-pixel convolution as up-sampling operations. Experimental results have indicated that our approaches, Kattolab, Kattolabv2 and KattolabSSIM, achieve 0.972 in MS-SSIM at the rate constraint of 0.15bpp with moderate complexity during the validation phase.