Unsupervised single image deraining with self-supervised constraints
Introduction
Most existing single image deraining methods require learn-ing supervised models from a large set of paired synthetictraining data, which limits their generality and practicalityin real-world multimedia applications. Besides, due to lackof labeled-supervised constraints, directly applying existingunsupervised frameworks to the image deraining task willsuffer from low-quality recovery. Therefore, we proposean Unsupervised Deraining Generative Adversarial Network(UD-GAN) to tackle above problems by introducing self-supervised constraints from the intrinsic statistics of unpairedrainy and clean images. Specifically, we design two collabo-ratively optimized modules, namely Rain Guidance Module(RGM) and Background Guidance Module (BGM), to takefull advantage of rainy image characteristics. UD-GAN out-performs state-of-the-art methods on various benchmarkingdatasets in both quantitative and qualitative comparisons.
Paper
Xin Jin, Zhibo Chen*, Jianxin Lin, Zhikai Chen, Wei Zhou. Unsupervised single image deraining with self-supervised constraints[C]//2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019: 2761-2765.
Download
To get the source code, Please fill THIS FORM and the source code will be sent to you.