Neural Image Compression via Non-Local Attention Optimization and Improved Context Modeling

TL;DR

This paper introduces a novel neural image compression method that leverages non-local attention and improved context modeling to achieve state-of-the-art performance in image compression, effectively capturing both local and global correlations.

Contribution

It presents a new deep neural network-based image compression framework integrating non-local operations, attention mechanisms, and advanced entropy modeling for improved efficiency.

Findings

Outperforms existing methods on Kodak and CLIC datasets

Achieves state-of-the-art compression efficiency in PSNR and MS-SSIM metrics

Enhances processing speed and reduces memory usage through novel techniques

Abstract

This paper proposes a novel Non-Local Attention optmization and Improved Context modeling-based image compression (NLAIC) algorithm, which is built on top of the deep nerual network (DNN)-based variational auto-encoder (VAE) structure. Our NLAIC 1) embeds non-local network operations as non-linear transforms in the encoders and decoders for both the image and the latent representation probability information (known as hyperprior) to capture both local and global correlations, 2) applies attention mechanism to generate masks that are used to weigh the features, which implicitly adapt bit allocation for feature elements based on their importance, and 3) implements the improved conditional entropy modeling of latent features using joint 3D convolutional neural network (CNN)-based autoregressive contexts and hyperpriors. Towards the practical application, additional enhancements are also introduced to speed up processing (e.g., parallel 3D CNN-based context prediction), reduce memory consumption (e.g., sparse non-local processing) and alleviate the implementation complexity (e.g., unified model for variable rates without re-training). The proposed model outperforms existing methods on Kodak and CLIC datasets with the state-of-the-art compression efficiency reported, including learned and conventional (e.g., BPG, JPEG2000, JPEG) image compression methods, for both PSNR and MS-SSIM distortion metrics.