Description

Analysing images, to estimate the underlying parameters that lead to their formation, is fundamentally an inverse problem. Since the observed image alone is usually not enough to uniquely determine these parameters, statistical models are frequently used to choose a likely solution from amongst those that are consistent with this observation. Deblurring is the process of removing blurring artifacts from images, such as blur caused by defocus aberration or motion blur. Motion blur is the apparent streaking of rapidly moving objects in a still image. A Gaussian blur is the result of blurring an image by a Gaussian function. A learning-based method of estimating blur kernel under the lo regularization sparsity constraint is proposed. lo Sparse representation to greatly benefit kernel estimation and large-scale optimization. The success of recent single-image methods partly stems from the use of various sparse priors, for either the latent images or motion blur kernels. Apply difference of convex functions programming which is a generic and principled way for solving non-smooth and non-convex optimization problem. Kernel Sparse Representation (KSR) is a sparse coding technique in a high dimensional feature space mapped by an implicit mapping function. kernel similarities without the knowledge of the actual form of the kernel, several pattern analysis and machine learning methods can be tractably solved using kernel methods. By choosing appropriate kernel functions that can extract task-specific information from the data, the learning problem can be effectively regularized. KSR also finds good kernel matrix approximation to speed up blurring and achieve good deblur performances on digital datasets.