The Study Of Image Restoration Based On Boundary Operator

In the past decades,the image restoration has been the emphasis of research in the field of computer and digital image processing.The main purpose of Image Reconstruction is to improve the blurred image with a predetermined goal.The technology of image restoration is for its degradation model,then use the reverse process to recover the original image.The convolutional sparse coding is an important tool in the field of the machine learning and computer vision.Instand of traditional sparse coding,convolutional sparse coding can operates the whole image in a convoluted form,thereby seamlessly capturing the correlation between local neighborhoods.Our paper has studied the image restoration based on the convolutional sparse coding and blind image restoration under the condition of the point spread function unknown.At the beginning,our paper reviews the history of convolutional sparse coding and image restoration.At the same time,the current research status and mathematical model also is mentioned.Then,based on image restoration with convolutional sparse coding and blind image restoration,a new method is proposed in our paper to deal with the questions below.Our paper choose the ADMM,based on separation variable as the main method to solve the questions.The background of ADMM also is introduced in our paper.The work is summarized below:First,focused on the problem of image restoration based on convolutional sparse coding.Our paper proposed a method with the aid of M matrix operator to deal with the boundary problem of image.M can be a binary diagonal matrix that masks out the boundaries of the padded estimation.This allow us to use unmodified filters in boundary regions,thus preserving the convolutional nature of the problem without requiring circular boundaries or other conditions.Our paper use the ADMM to solve the original problem.The use of ADMM can greatly accelerate iteration times.The final experimental result show that our method for image restoration based on convolutional sparse coding is effective and feasible.Second,focused on the problem of blind image restoration,our paper proposed a method,based on traditional TV model,combines the advantages of sparse regularization terms.Our method can suppress the ill-conditioned of image better.The pixels located near the boundary of the blurred image will tend to 0 with the help of the M matrix operator.The M matrix operator can prevent the degradation of the restored images.The final results show that our method for blind image restoration is effective and feasible.