Bidirectional Stereo Image Compression with Cross-Dimensional Entropy Model

TL;DR

This paper introduces BiSIC, a symmetric bidirectional stereo image compression method utilizing 3D convolutions and a novel cross-dimensional entropy model, significantly improving compression quality over existing standards.

Contribution

The paper presents a new symmetric bidirectional stereo image compression architecture with a cross-dimensional entropy model, addressing imbalance issues in prior unidirectional methods.

Findings

Outperforms traditional image/video compression standards

Achieves higher PSNR and MS-SSIM scores

Effective in capturing local and global features

Abstract

With the rapid advancement of stereo vision technologies, stereo image compression has emerged as a crucial field that continues to draw significant attention. Previous approaches have primarily employed a unidirectional paradigm, where the compression of one view is dependent on the other, resulting in imbalanced compression. To address this issue, we introduce a symmetric bidirectional stereo image compression architecture, named BiSIC. Specifically, we propose a 3D convolution based codec backbone to capture local features and incorporate bidirectional attention blocks to exploit global features. Moreover, we design a novel cross-dimensional entropy model that integrates various conditioning factors, including the spatial context, channel context, and stereo dependency, to effectively estimate the distribution of latent representations for entropy coding. Extensive experiments demonstrate that our proposed BiSIC outperforms conventional image/video compression standards, as well as state-of-the-art learning-based methods, in terms of both PSNR and MS-SSIM.