Abstract:It is well-known that shaken cameras or mobile phones during exposure usually lead to motion blurry photographs. Therefore, camera shake deblurring or motion deblurring is required and requested in many practical scenarios. The contribution of this paper is the proposal of a simple yet effective approach for motion blur kernel estimation, i.e., blind motion deblurring. Though there have been proposed several methods for motion blur kernel estimation in the literature, we impose a type of non-stationary Gaussian prior on the gradient fields of sharp images, in order to automatically detect and purse the salient edges of images as the important clues to blur kernel estimation. On one hand, the prior is able to promote sparsity inherited in the non-stationarity of the precision parameters (inverse of variances). On the other hand, since the prior is in a Gaussian form, there exists a great possibility of deducing a conceptually simple and computationally tractable inference scheme. Specifically, the well-known expectation–maximization algorithm is used to alternatingly estimate the motion blur kernels, the salient edges of images as well as the precision parameters in the image prior. In difference from many existing methods, no hyperpriors are imposed on any parameters in this paper; there are not any pre-processing steps involved in the proposed method, either, such as explicit suppression of random noise or prediction of salient edge structures. With estimated motion blur kernels, the deblurred images are finally generated using an off-the-shelf non-blind deconvolution method proposed by Krishnan and Fergus (Adv Neural Inf Process Syst 22:1033–1041, 2009). The rationality and effectiveness of our proposed method have been well demonstrated by the experimental results on both synthetic and realistic motion blurry images, showing state-of-the-art blind motion deblurring performance of the proposed approach in the term of quantitative metric as well as visual perception.

Efficient learning-based blur removal method based on sparse optimization for image restoration

Image Deblurring Utilizing Inertial Sensors and a Short-Long-Short Exposure Strategy.

Image Deblurring Using Robust Sparsity Priors

Estimation of blur parameters in image restoration

Learning to Estimate and Remove Non-uniform Image Blur

Kernel Learning for Blind Image Recovery from Motion Blur

Regularized Motion Blur-Kernel Estimation with Adaptive Sparse Image Prior Learning.

Blur Kernel Estimation Using Sparse Representation and Cross-Scale Self-Similarity

Motion Deblurring from a Single Image Using Gradient Enhancement.

Blind Infrared Remote-Sensing Image Deblurring Algorithm via Edge Composite-Gradient Feature Prior and Detail Maintenance

Multi-Regularization-Constrained Blur Kernel Estimation Method for Blind Motion Deblurring

Motion Blurred Images Restoration Based on Sparse Regularization

Robust motion blur kernel parameter estimation for star image deblurring

Sparse Representation of a Blur Kernel for Blind Image Restoration

A Self-Adaptive Learning Method for Motion Blur Kernel Estimation of the Single Image

Motion Blur Kernel Estimation Via Deep Learning

Parametric Blur Estimation for Blind Restoration of Atmospherically Degraded Images: Class G

Natural Image Deblurring Based On Ringing Artifacts Removal Via Knowledge-Driven Gradient Distribution Priors

Improved estimation of motion blur parameters for restoration from a single image

Motion Deblurring Using Non-stationary Image Modeling

Efficient Patch-Wise Non-Uniform Deblurring for a Single Image