OMR-NET: a two-stage octave multi-scale residual network for screen content image compression

TL;DR

This paper introduces OMR-NET, a novel two-stage multi-scale residual network tailored for screen content image compression, effectively capturing high-contrast and repetitive patterns to outperform existing learned image compression methods.

Contribution

It proposes a new two-stage octave convolutional residual architecture with attention mechanisms and a diverse SC dataset, specifically optimized for screen content image compression.

Findings

Outperforms existing LIC methods in rate-distortion on SC images.

Effective multi-scale learning improves compression quality.

New dataset enhances training for screen content images.

Abstract

Screen content (SC) differs from natural scene (NS) with unique characteristics such as noise-free, repetitive patterns, and high contrast. Aiming at addressing the inadequacies of current learned image compression (LIC) methods for SC, we propose an improved two-stage octave convolutional residual blocks (IToRB) for high and low-frequency feature extraction and a cascaded two-stage multi-scale residual blocks (CTMSRB) for improved multi-scale learning and nonlinearity in SC. Additionally, we employ a window-based attention module (WAM) to capture pixel correlations, especially for high contrast regions in the image. We also construct a diverse SC image compression dataset (SDU-SCICD2K) for training, including text, charts, graphics, animation, movie, game and mixture of SC images and NS images. Experimental results show our method, more suited for SC than NS data, outperforms existing LIC methods in rate-distortion performance on SC images. The code is publicly available at https://github.com/SunshineSki/OMR Net.git.