Image Priors for Image Deblurring with Uncertain Blur
[1] Universität Bern, Bern, Switzerland
[2] UCLA VisionLab, University of California, Los Angeles, CA, USA
[3] Center for Imaging Science, Johns Hopkins University, Baltimore, MD, USA
Main | Comparison with Cho et al. | Comparison with Xu et al. | Levin Dataset |
Abstract
We consider the problem of non-blind deconvolution of images corrupted by a blur that is not accurately known. We propose a method that exploits dictionary-based image priors and non Gaussian noise models to improve deblurring accuracy in the presence of an inexact blur. The proposed image priors express each image patch as a linear combination of atoms from a dictionary learned from patches extracted from the same image or from an image database. When applied to blurred images, this model imposes that patches that are similar in the blurred image retain the same similarity when deblurred. We perform image deblurring by imposing this prior model in an energy minimization scheme that also deals with outliers. Experimental results on publicly available databases show that our approach is able to remove artifacts such as oscillations, which are often introduced during the deblurring process when the correct blur is not known.
Results
We performed several comparisons with publicly available examples. Please use the links in the navigation bar to see the results for different datasets and examples.
Bibtex
@inproceedings{perrone_bmvc2012,
author = {Daniele Perrone, Avinash Ravichandran, René Vidal and Paolo Favaro},
title = {Image Priors for Image Deblurring with Uncertain Blur},
booktitle = {Proceedings of the British Machine Vision Conference},
year = {2012}
}
Downloads
Paper.Acknowledgments
Daniele Perrone and Paolo Favaro have been supported by grant ONR N00014-09-1-1067, Google Research Award 113095 and SELEX/HWU/2010/SOW4 from Selex-Galileo. Avinash Ravichandran and René Vidal have been supported by grant ONR N00014-09-10084.