Research On Image Deblurring Algorithm

Blurring artifacts are the most common flaws in photographs,which could be caused by the movement during the exposure or the defocus of the camera.How to recover the latent image from the blurred image is an important issue in the low-level image processing.The blur caused by defocusing or moving objects is generally spatially-varying.And there are different degrees of blur in different areas of the blurred image,which make the deblurring problem more difficult to solve.Aiming at the problem of defocus blur,an image defocus estimation and removal algorithm based on significant step edge structure is proposed.In order to reduce the influence of defocused texture areas and smooth areas,the L0 gradient sparse prior is used to estimate the significant step edge structure and the defocus variance is estimated at the edge of the significant step edge structure.Thereby the influence of non-step edges and smooth areas on the estimation of the defocus map is reduced,and the accuracy of the defocus map is improved.Experiments show that the proposed method can reduce the mean square error of the defocused image on both the synthetic defocused image and the natural defocused image.The average peak signal-to-noise ratio of the clear image obtained on the synthetic defocus image without foreground can be increased by 0.203 d B,and the average peak signal-to-noise ratio of the synthetic defocus image with foreground can be increased by 0.590 d B.Aiming at the blur problem of moving object,the motion blur estimation and removal method of local linear motion blur is proposed by taking advantage of the characteristics of moving objects in local areas with similar motion directions.By dividing the motion region into blocks and applying linear motion blur constraints to each image block,the blur kernel corresponding to the region is solved,so as to obtain the blur matrix corresponding to the entire motion region.Finally,the method of nonblindly solving spatially-variable convolution is used to estimate the clear image.Experiments show that the deblurring method for moving objects based on local linear blurring has an average increase of 2.105 d B in peak signal-to-noise ratio and an average increase of 0.034 in structural similarity compared with other deblurring methods based on optimization.Finally,by combining the defocus estimation and removal algorithm based on the significant step edge and the object motion deblurring algorithm based on the local linear constraint,it is used to process the scene where there are both moving object blur and defocus blur.The blur matrices of defocus and motion blur is used to establish the objective function to solve the clear image,and the clear image when there is mixed blur is obtained through optimization.