Paper Info
Title
Learning Frequency Domain Priors for Image Demoireing
Abstract
Image demoireing is a multi-faceted image restoration task involving both moire pattern removal and color restoration. In this paper, we raise a general degradation model to describe an image contaminated by moire patterns, and propose a novel multi-scale bandpass convolutional neural network (MBCNN) for single image demoireing. For moire pattern removal, we propose a multi-block-size learnable bandpass filters (M-LBFs), based on a block-wise frequency domain transform, to learn the frequency domain priors of moire patterns. We also introduce a new loss function named Dilated Advanced Sobel loss (D-ASL) to better sense the frequency information. For color restoration, we propose a two-step tone mapping strategy, which first applies a global tone mapping to correct for a global color shift, and then performs local fine tuning of the color per pixel. To determine the most appropriate frequency domain transform, we investigate several transforms including DCT, DFT, DWT, learnable non-linear transform and learnable orthogonal transform. We finally adopt the DCT. Our basic model won the AIM2019 demoireing challenge. Experimental results on three public datasets show that our method outperforms state-of-the-art methods by a large margin.
Year
DOI
Venue
2022
10.1109/TPAMI.2021.3115139
IEEE Transactions on Pattern Analysis and Machine Intelligence
Keywords
DocType
Volume
Image demoireing,frequency domain prior,learnable bandpass filter,dilated advanced sobel loss,degradation model,learnable orthogonal transform,two-step color restoration
Journal
44
Issue
ISSN
Citations 
11
0162-8828
2
PageRank 
References 
Authors
0.36
21
9
Name
Order
Citations
PageRank
Bolun Zheng121.37
Shanxin Yuan2355.38
Chenggang Yan313316.62
Xiang Tian4154.77
Jiyong Zhang515621.11
Yaoqi Sun640.74
Lin Liu7136.35
Ales Leonardis81636147.33
Gregory G. Slabaugh987071.13