Single Image Super-Resolution via a Holistic Attention Network

TL;DR

This paper introduces a holistic attention network (HAN) for single image super-resolution that models interdependencies among layers, channels, and spatial positions to enhance feature extraction and improve reconstruction quality.

Contribution

The paper proposes a novel HAN architecture with layer and channel-spatial attention modules to better capture feature correlations across layers and spatial positions.

Findings

HAN outperforms existing super-resolution methods in experiments

The layer attention module effectively emphasizes hierarchical features

The channel-spatial attention module captures more informative features

Abstract

Informative features play a crucial role in the single image super-resolution task. Channel attention has been demonstrated to be effective for preserving information-rich features in each layer. However, channel attention treats each convolution layer as a separate process that misses the correlation among different layers. To address this problem, we propose a new holistic attention network (HAN), which consists of a layer attention module (LAM) and a channel-spatial attention module (CSAM), to model the holistic interdependencies among layers, channels, and positions. Specifically, the proposed LAM adaptively emphasizes hierarchical features by considering correlations among layers. Meanwhile, CSAM learns the confidence at all the positions of each channel to selectively capture more informative features. Extensive experiments demonstrate that the proposed HAN performs favorably against the state-of-the-art single image super-resolution approaches.