Neural Distributed Image Compression with Cross-Attention Feature Alignment

TL;DR

This paper introduces a neural image compression method that uses cross-attention to align features from stereo images, effectively utilizing side information at the decoder for improved compression performance.

Contribution

It proposes a novel cross-attention feature alignment module in neural image compression that better exploits decoder-only side information, advancing distributed source coding techniques.

Findings

Achieves competitive compression performance on KITTI and Cityscape datasets.

Demonstrates improved utilization of side information compared to previous methods.

Shows that feature alignment enhances the efficiency of distributed image compression.

Abstract

We consider the problem of compressing an information source when a correlated one is available as side information only at the decoder side, which is a special case of the distributed source coding problem in information theory. In particular, we consider a pair of stereo images, which have overlapping fields of view, and are captured by a synchronized and calibrated pair of cameras as correlated image sources. In previously proposed methods, the encoder transforms the input image to a latent representation using a deep neural network, and compresses the quantized latent representation losslessly using entropy coding. The decoder decodes the entropy-coded quantized latent representation, and reconstructs the input image using this representation and the available side information. In the proposed method, the decoder employs a cross-attention module to align the feature maps obtained from the received latent representation of the input image and a latent representation of the side information. We argue that aligning the correlated patches in the feature maps allows better utilization of the side information. We empirically demonstrate the competitiveness of the proposed algorithm on KITTI and Cityscape datasets of stereo image pairs. Our experimental results show that the proposed architecture is able to exploit the decoder-only side information in a more efficient manner compared to previous works.