Adaptive Integration Skip Compensation Neural Networks for Removing Mixed Noise in Image

During the process of acquisition and transmission, images are often likely to be corrupted by mixed Gaussian-impulse noise. Among various image denoising methods, most traditional methods can only deal with a single type of noise due to the difficulty of modeling the distribution of the mixed noise. In this paper, we propose a novel mixed Gaussian-impulse noise removal method based on adaptive integration skip compensation Network (Ai-Sc-Net). More concretely, a couple of skip compensation networks (Sc-Net) Sc-Net-AWGN and Sc-Net-IN are trained on Gaussian and Impulse noise datasets separately to deal with the corresponding single type noise. Further, an adaptive integration network (Ai-Net) is used to integrate the two outputs of Sc-Net-AWGN and Sc-Net-IN. The Ai-Sc-Net is then be constructed based on Sc-Net and Ai-Net, which can handle mixed noise. Experimental results in synthetic noise images have shown great improvements over several state-of-the-art mixed noise removal methods.

[1]  Roberto Manduchi,et al.  Bilateral filtering for gray and color images , 1998, Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271).

[2]  Charles K. Chui,et al.  A universal noise removal algorithm with an impulse detector , 2005, IEEE Transactions on Image Processing.

[3]  Alan C. Bovik,et al.  Image information and visual quality , 2006, IEEE Trans. Image Process..

[4]  Qingjin Peng,et al.  Robust recognition of checkerboard pattern for camera calibration , 2006 .

[5]  Alessandro Foi,et al.  Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering , 2007, IEEE Transactions on Image Processing.

[6]  H. Sebastian Seung,et al.  Natural Image Denoising with Convolutional Networks , 2008, NIPS.

[7]  Pascal Vincent,et al.  Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion , 2010, J. Mach. Learn. Res..

[8]  Jian Yu,et al.  Restoration of images corrupted by mixed Gaussian-impulse noise via l1-l0 minimization , 2011, Pattern Recognit..

[9]  Ming Yan,et al.  Restoration of Images Corrupted by Impulse Noise and Mixed Gaussian Impulse Noise using Blind Inpainting , 2013, SIAM J. Imaging Sci..

[10]  Lei Zhang,et al.  Nonlocally Centralized Sparse Representation for Image Restoration , 2013, IEEE Transactions on Image Processing.

[11]  Jian Yang,et al.  Mixed Noise Removal by Weighted Encoding With Sparse Nonlocal Regularization , 2014, IEEE Transactions on Image Processing.

[12]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[13]  Yu-Bin Yang,et al.  Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections , 2016, NIPS.

[14]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[15]  Lei Zhang,et al.  Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising , 2016, IEEE Transactions on Image Processing.

[16]  Xiang Huang,et al.  Adaptive Bregmanized total variation model for mixed noise removal , 2017 .

[17]  Guangming Shi,et al.  Mixed Noise Removal via Laplacian Scale Mixture Modeling and Nonlocal Low-Rank Approximation , 2017, IEEE Transactions on Image Processing.

[18]  Yuan Yan Tang,et al.  Weighted Joint Sparse Representation for Removing Mixed Noise in Image , 2017, IEEE Transactions on Cybernetics.

[19]  Gangyao Kuang,et al.  Hyperspectral Image Restoration Using Low-Rank Tensor Recovery , 2017, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[20]  Luca Antiga,et al.  Automatic differentiation in PyTorch , 2017 .

[21]  Yunjin Chen,et al.  Trainable Nonlinear Reaction Diffusion: A Flexible Framework for Fast and Effective Image Restoration , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.