Rethinking Performance Gains in Image Dehazing Networks

TL;DR

This paper introduces a minimal modification to the U-Net architecture, called gUNet, which uses residual blocks with gating and selective kernel fusion to improve image dehazing performance efficiently.

Contribution

The paper proposes a simplified yet effective U-Net variant for image dehazing, demonstrating that key architectural modifications can lead to significant performance gains.

Findings

gUNet outperforms state-of-the-art methods on multiple datasets

Key design choices are validated through extensive ablation studies

Achieves better performance with reduced computational overhead

Abstract

Image dehazing is an active topic in low-level vision, and many image dehazing networks have been proposed with the rapid development of deep learning. Although these networks' pipelines work fine, the key mechanism to improving image dehazing performance remains unclear. For this reason, we do not target to propose a dehazing network with fancy modules; rather, we make minimal modifications to popular U-Net to obtain a compact dehazing network. Specifically, we swap out the convolutional blocks in U-Net for residual blocks with the gating mechanism, fuse the feature maps of main paths and skip connections using the selective kernel, and call the resulting U-Net variant gUNet. As a result, with a significantly reduced overhead, gUNet is superior to state-of-the-art methods on multiple image dehazing datasets. Finally, we verify these key designs to the performance gain of image dehazing networks through extensive ablation studies.