Cycle-Dehaze: Enhanced CycleGAN for Single Image Dehazing

Engin D., Genc A., Ekenel H. K.

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Utah, United States Of America, 18 - 22 June 2018, pp.938-946 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/cvprw.2018.00127
  • City: Utah
  • Country: United States Of America
  • Page Numbers: pp.938-946
  • Istanbul Technical University Affiliated: Yes


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 down-scaling. 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.