Attention-guided Image Compression by Deep Reconstruction of Compressive Sensed Saliency Skeleton

TL;DR

This paper introduces a deep learning-based attention-guided dual-layer image compression system that selectively compresses perceptually critical pixels using CNNs and compressive sensing, achieving superior perceptual quality.

Contribution

It presents a novel AGDL system that predicts and compresses critical pixels within ROI, combining CNNs and compressive sensing for improved perception-aware image compression.

Findings

Outperforms existing ROI compression methods in perceptual quality.

Effectively predicts and encodes critical pixels using CNN and CS.

Achieves refined image reconstruction with joint decoding of two layers.

Abstract

We propose a deep learning system for attention-guided dual-layer image compression (AGDL). In the AGDL compression system, an image is encoded into two layers, a base layer and an attention-guided refinement layer. Unlike the existing ROI image compression methods that spend an extra bit budget equally on all pixels in ROI, AGDL employs a CNN module to predict those pixels on and near a saliency sketch within ROI that are critical to perceptual quality. Only the critical pixels are further sampled by compressive sensing (CS) to form a very compact refinement layer. Another novel CNN method is developed to jointly decode the two compression layers for a much refined reconstruction, while strictly satisfying the transmitted CS constraints on perceptually critical pixels. Extensive experiments demonstrate that the proposed AGDL system advances the state of the art in perception-aware image compression.