A Neural Approach to Blind Motion Deblurring

We present a new method for blind motion deblurring that uses a neural network trained to compute estimates of sharp image patches from observations that are blurred by an unknown motion kernel. Instead of regressing directly to patch intensities, this network learns to predict the complex Fourier coefficients of a deconvolution filter to be applied to the input patch for restoration. For inference, we apply the network independently to all overlapping patches in the observed image, and average its outputs to form an initial estimate of the sharp image. We then explicitly estimate a single global blur kernel by relating this estimate to the observed image, and finally perform non-blind deconvolution with this kernel. Our method exhibits accuracy and robustness close to state-of-the-art iterative methods, while being much faster when parallelized on GPU hardware.

[1]  Rob Fergus,et al.  Restoring an Image Taken through a Window Covered with Dirt or Rain , 2013, 2013 IEEE International Conference on Computer Vision.

[2]  Frédo Durand,et al.  Understanding and evaluating blind deconvolution algorithms , 2009, 2009 IEEE Conference on Computer Vision and Pattern Recognition.

[3]  Luc Van Gool,et al.  The Pascal Visual Object Classes Challenge: A Retrospective , 2014, International Journal of Computer Vision.

[4]  Guanrong Chen,et al.  Introduction to random signals and applied Kalman filtering, 2nd edn. Robert Grover Brown and Patrick Y. C. Hwang, Wiley, New York, 1992. ISBN 0‐471‐52573‐1, 512 pp., $62.95. , 1992 .

[5]  Bernhard Schölkopf,et al.  A Machine Learning Approach for Non-blind Image Deconvolution , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[6]  Donald Geman,et al.  Nonlinear image recovery with half-quadratic regularization , 1995, IEEE Trans. Image Process..

[7]  Jean Ponce,et al.  Learning a convolutional neural network for non-uniform motion blur removal , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[9]  Rob Fergus,et al.  Blind deconvolution using a normalized sparsity measure , 2011, CVPR 2011.

[10]  Li Xu,et al.  Two-Phase Kernel Estimation for Robust Motion Deblurring , 2010, ECCV.

[11]  Jan Kotera,et al.  Convolutional Neural Networks for Direct Text Deblurring , 2015, BMVC.

[12]  Ce Liu,et al.  Deep Convolutional Neural Network for Image Deconvolution , 2014, NIPS.

[13]  Frédo Durand,et al.  Efficient marginal likelihood optimization in blind deconvolution , 2011, CVPR 2011.

[14]  Sylvain Paris,et al.  Blur kernel estimation using the radon transform , 2011, CVPR 2011.

[15]  Sunghyun Cho,et al.  Edge-based blur kernel estimation using patch priors , 2013, IEEE International Conference on Computational Photography (ICCP).

[16]  Yair Weiss,et al.  From learning models of natural image patches to whole image restoration , 2011, 2011 International Conference on Computer Vision.

[17]  Seungyong Lee,et al.  Fast motion deblurring , 2009, ACM Trans. Graph..

[18]  Sundaresh Ram,et al.  Removing Camera Shake from a Single Photograph , 2009 .

[19]  Bernhard Schölkopf,et al.  Learning to Deblur , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[20]  Stefan Harmeling,et al.  Image denoising: Can plain neural networks compete with BM3D? , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[21]  Michal Irani,et al.  Blind Deblurring Using Internal Patch Recurrence , 2014, ECCV.

[22]  Klaus-Robert Müller,et al.  Efficient BackProp , 2012, Neural Networks: Tricks of the Trade.