Efficient marginal likelihood optimization in blind deconvolution

By Anat Levin, Yair Weiss, Frédo Durand, William T. Freeman. Computer Vision and Pattern Recognition, 2011

666 citations recorded by the source corpus.

In blind deconvolution one aims to estimate from an input blurred image y a sharp image x and an unknown blur kernel k. Recent research shows that a key to success is to consider the overall shape of the posterior distribution p(x, k\y) and not only its mode. This leads to a distinction between MAPx, k strategies which estimate the mode pair x, k and often lead to undesired results, and MAPk strategies which select the best k while marginalizing over all possible x images. The MAPk principle is significantly more robust than the MAPx, k one, yet, it involves a challenging marginalization over latent images. As a result, MAPk techniques are considered complicated, and have not been widely exploited. This paper derives a simple approximated MAPk algorithm which involves only a modest modification of common MAPx, k algorithms. We show that MAPk can, in fact, be optimized easily, with no additional computational complexity.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Signal Processing

View Paper: https://www.semanticscholar.org/paper/a0925c6da392a624eefa78cd3974ba5b00f57cd9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

In Signal Processing

Citations in the corpus

A Neural Approach to Blind Motion Deblurring (2016, methodology): “[1], the last decade has seen considerable progress [2,3,4,5,6,7,8] in the development of effective blind motion deblurring methods that seek to estimate camera motion in terms of the induced blur kernel, and then reverse its effect.”
Total Variation Blind Deconvolution: The Devil Is in the Details (2014, background): “However, presentday progress has shown that recent models for sharp images and blur kernels, such as total variation [18], can yield remarkable results [7, 19, 4, 23, 13].”
A Blur-Robust Descriptor with Applications to Face Recognition (2012, background): “…studied by the image processing community over the last four decades [4], and some of the prominent methodologies include: blind deconvolution [20] that does not assume any knowledge of the blur kernel and thereby attempts to solve an under-constrained problem of estimating both k and y from…”
Hand-Held Video Deblurring Via Efficient Fourier Aggregation (2015, background): “Due to its small spatial support, the blurring kernel estimation is an easier problem to solve than simultaneously estimating both the kernel and the sharp image [21], [22].”
Deblurring Text Images via L0-Regularized Intensity and Gradient Prior (2014, methodology): “For each sharp image, we compute the average PSNR on the blurred images from different kernels and compare among different methods [16, 3, 18, 10, 12, 20, 21] in Figure 8.”
Unnatural L0 Sparse Representation for Natural Image Deblurring (2013, methodology): “[18] (Matlab) 76.”
Nonparametric Blind Super-resolution (2013, methodology): “However, unlike [14], our MAPk is based on a nonparametric prior.”
Blind Image Deblurring Using Dark Channel Prior (2016, methodology): “We note that kernel estimation methods based on gradients have been shown to be more accurate [5, 20, 36] (see analysis in the supplemental material).”
From Motion Blur to Motion Flow: A Deep Learning Solution for Removing Heterogeneous Motion Blur (2016, methodology): “Various estimators have been proposed for more robust kernel estimation, such as edge-extraction-based maximum-a-posteriori (MAP) [6, 34], gradient activation based MAP [10], variational Bayesian methods [21, 22, 46], etc .”
Blind Deblurring Using Internal Patch Recurrence (2014, methodology): “Some studies assume simple parametric probability models, which promote sparsity of image gradients [5,17,12,13,10].”
A no-reference metric for evaluating the quality of motion deblurring (2013, methodology): “We run five recent algorithms [Fergus et al. 2006; Shan et al. 2008; Cho and Lee 2009; Levin et al. 2011; Goldstein and Fattal 2012] with publicly-available executables on all the blurred images to generate deblurring results.”
$L_0$ -Regularized Intensity and Gradient Prior for Deblurring Text Images and Beyond (2017, methodology): “14 shows one example with numerous saturated regions and deblurred results from the state-of-the-art methods [3], [4], [5], [6], [7], [47], [48].”
Image Deblurring via Extreme Channels Prior (2017, methodology): “To estimate blur kernels from blurry images, other existing approaches utilize statistical priors[4, 20, 14, 12, 25, 16] or additional information [8, 3, 9] to solve the ill-posed problem.”
A Comparative Study for Single Image Blind Deblurring (2016, neutral): “We evaluate 13 representative state-of-the-art algorithms [2] [5] [15] [18] [21] [29] [31] [38] [43] [44] [45] [48] [49] in our experiments 1 .”