Image Deblurring Based On Local Adaptive Sparsity Constraint

With the rapid development of digital technology, people are contact with digital images constantly in the life and work. In the process of image acquisition, the relative motion between equipment and scene may cause blur in the images, and the blur can give people a lot of inconvenience. In recent years, the issue of image deblurring has become a hot research topic in the field of image processing in the domestic and overseas. Image restoration technology is also widely used in the field of public security, medical imaging, traffic safety surveillance, military reconnaissance, aerospace, etc. Therefore, the subject has a certain research significance and important application value.This paper focuses on a single image blind motion deblurring problem. Firstly, based on the research background of image deblurring, we emphasize its important research ignificance and application value, and expound the advantages and deficiencies of the existing methods. Secondly, we analyze the ill posedness of the deblurring problem and briefly explain the solving method based on regularization. According to the priori information of the natural image gradient and kernel, we introduce several typical corresponding sparse prior distributions and sparse constraint functions.In this paper, using the local information of image, a regularization method based on local adaptive sparse constraint is proposed to adjust the penalty in different regions of the image adaptively. The method can automatically reduce the punishment of the edge region, and increase the punishment of the flat region, which makes the solution of the model tending to the clear image.As the image salient structure has an important influence on the accuracy of kernel estimation, we extract the image structure based on the RTV method, and combine the local adaptive sparse constraint regularization model to estimate kernel with multi scale method. In addition, in order to suppress the noise in the kernel and maintain the continuity and sparse of kernel structure, we use the effective method based on l0norm to refine the preliminary kernel estimation.Experimental results show that the method of this paper can guarantee sparsity and continuity of the kernel. At the same time, the method can effectively remove the blur and noise in image, effectively suppress the ringing, maintain the sharpness of image edge and obtain a high-quality image restoration finally.