Cycle-Dehaze: Enhanced CycleGAN for Single Image Dehazing

TL;DR

Cycle-Dehaze introduces an unpaired, end-to-end deep learning model that enhances image dehazing quality without ground truth pairs, utilizing cycle-consistency, perceptual losses, and multi-scale upscaling techniques.

Contribution

It advances dehazing methods by combining cycleGAN with perceptual loss and multi-scale upscaling, eliminating the need for paired training data and atmospheric model estimation.

Findings

Improves dehazing quality both quantitatively and qualitatively.

Effective on high-resolution images using Laplacian pyramid upscaling.

Outperforms previous cycleGAN-based dehazing methods.

Abstract

In this paper, we present an end-to-end network, called Cycle-Dehaze, for single image dehazing problem, which does not require pairs of hazy and corresponding ground truth images for training. That is, we train the network by feeding clean and hazy images in an unpaired manner. Moreover, the proposed approach does not rely on estimation of the atmospheric scattering model parameters. Our method enhances CycleGAN formulation by combining cycle-consistency and perceptual losses in order to improve the quality of textural information recovery and generate visually better haze-free images. Typically, deep learning models for dehazing take low resolution images as input and produce low resolution outputs. However, in the NTIRE 2018 challenge on single image dehazing, high resolution images were provided. Therefore, we apply bicubic downscaling. After obtaining low-resolution outputs from the network, we utilize the Laplacian pyramid to upscale the output images to the original resolution. We conduct experiments on NYU-Depth, I-HAZE, and O-HAZE datasets. Extensive experiments demonstrate that the proposed approach improves CycleGAN method both quantitatively and qualitatively.