Research On Image Restoration Algorithm Based On Single Motion Blurred Image

Along with the advance of computer vision technology, the quality of the images more and more get people's attention. It is a significant discussion topic between scholars that how can effectively restore the degraded images. The problem of the motion blur image restoration is one of the hot issues, which are widely used in the fields of astronomy, military, transportation,medicine and daily life. It has many difficulties to restore the high quality image because of some unknown variables and noise information.This paper is mainly studied the motion blurred image restoration of unknown blurring kernel from the perspective of a single motion blurred image. The main research content includes the following two aspects:Being aimed at the problem of image blurred kernel, this paper proposes the kernel estimation algorithm based on L1 regularization method.First of all, the blurred kernel is estimated by the input blurred image. Secondly, the image is restored according to the estimated blurred kernel, and the blurred kernel continues to estimate by the restored images. Two steps are calculated alternately until the correct blurred kernel is obtained. Finally, the blurred kernel is corrected by the blurred kernel character. Experiments demonstrate that the proposed method not only has certain anti - noise, but also improves the accuracy of kernel estimation.By considering the problem of image restoration, this paper proposes an image restoration algorithm based on sparse prior and super-Laplace. Firstly, the motion blurred image is preprocessed by using the guided filter to remove the noise information, and the improved shock filter enhances the edge information of the image. Secondly, L1-regularization method is used to evaluate the blur kernel of the blurred image alternately. Finally, sparse priors and Hyper- Laplacian are applied to the image obtained from the non-blind deconvolution process.In this paper, the spilt-Bregman algorithm is used to improve the overall running time of the algorithm for the problem of L1 regularization optimal solution in image restoration algorithm.It can be observed in the experiment that the proposed method have obvious advantages in the effect and efficiency of image restoration.