An Effective Single-Image Super-Resolution Model Using Squeeze-and-Excitation Networks

TL;DR

This paper introduces a novel single-image super-resolution model that leverages squeeze-and-excitation blocks to adaptively recalibrate channel features, achieving state-of-the-art results with finer texture details.

Contribution

The paper proposes a deep residual network incorporating SEBlocks and short connections to improve super-resolution performance by modeling channel correlations.

Findings

Achieves state-of-the-art super-resolution performance.

Reconstructs finer texture details.

Outperforms existing methods in experiments.

Abstract

Recent works on single-image super-resolution are concentrated on improving performance through enhancing spatial encoding between convolutional layers. In this paper, we focus on modeling the correlations between channels of convolutional features. We present an effective deep residual network based on squeeze-and-excitation blocks (SEBlock) to reconstruct high-resolution (HR) image from low-resolution (LR) image. SEBlock is used to adaptively recalibrate channel-wise feature mappings. Further, short connections between each SEBlock are used to remedy information loss. Extensive experiments show that our model can achieve the state-of-the-art performance and get finer texture details.