Learning a convolutional neural network for non-uniform motion blur removal

Abstract

International audienceIn this paper, we address the problem of estimating and removing non-uniform motion blur from a single blurry image. We propose a deep learning approach to predicting the probabilistic distribution of motion blur at the patch level using a convolutional neural network (CNN). We further extend the candidate set of motion kernels predicted by the CNN using carefully designed image rotations. A Markov random field model is then used to infer a dense non-uniform motion blur field enforcing motion smoothness. Finally, motion blur is removed by a non-uniform de-blurring model using patch-level image prior. Experimental evaluations show that our approach can effectively estimate and remove complex non-uniform motion blur that is not handled well by previous approaches

Learning a Deep Convolutional Network for Light-Field Image Super-Resolution

“…Deep learning for image restoration Recently, the deep learning approach has been successfully applied to the image restoration problems such as SR [8,11], denoising [2,12,22] and image deblurring [25,32]. Schuler et al [22] show that a multi-layer perceptron can solve an image deconvolution problem.…”

“…The first network is for deconvolution and the second network is built based on [12] to remove outliers. Sun et al [25] design a CNN for non-uniform motion blur estimation with Markov Random Field.…”

“…There are different lines of work that adopt deep learning frameworks to solve image restoration problems, including SR [11,8], denoising [12] and deblurring [32,25]. Compared to them, our major focus lies on LF imaging, which is the specific domain having different restoration problems and applications.…”

A Robust Image Super-solution Algorithm Using Deep Convolution Networks

“…Deep learning also can be used to solve some low-level vision problems, for instance image deblurring [19], super-solution [20]. In this paper, on the basis of end-to-end learning framework using convolutional neural network [20], we put forward a robust SISR algorithm by exploring the architecture of convolutional neural network (CNN) and combining different network architectures.…”

Research on Blind Image Restoration Based on Two Dimensional Constant Modulus Algorithm

“…In other words, the blind image restoration technique restores the specific PSF and the specific original image only based on the "specific" image observation data. Obviously, blind image restoration algorithm is an image recovery method with more practical [4,5].In recent years, with popularity of computers and computer networks, blind image restoration research has gained unprecedented attention as a popular topic in signal processing. New techniques and methods such as parameter estimation method [6,7], non-parametric estimation method [8,9], and fast algorithm [10] were proposed.…”