Research On Motion Blurred Image Restoration Method Based On The Total Variation Model

The relative movement of the photographic material and the photographing device causes motion blurred image. Motion blurred image restoration is a hot research and has been widely used in video surveillance, traffic management, criminal investigation,satellite remote sensing and so on. Therefore, it is of great practical significance to improve the restoration of motion blurred image. This thesis focuses on the research of the total variation(TV) blind deconvolution method in image restoration, and improves the TV method from two aspects: priori blind deblurring and joint blind deconvolution.The main research contents are as follows.Firstly, a priori blind deblurring method is improved to solve the problem of inaccurate estimation of point spread function(PSF). It is a kind of PSF estimation method based on image gradient analysis and phase retrieval. In this method, the PSF energy information is obtained by using image gradient inverse spectral information.Then, the PSF phase is obtained by using the alternating fourier transform method.Finally, combined with the PSF energy information and PSF phase information to get the complete PSF. The experimental results show that the PSF obtained by this method is more similar to the real PSF. In order to solve the problem of non ideal reconstruction of the image edges and corners, the thesis uses a restoration method combining high order constraints and anisotropic TV. First, the estimated PSF is used as an input. Then,a restoration model is constructed by combining anisotropic TV and higher order constraints. At last, the model is used to restore the blurred image. The experimental results show that the proposed method has a better restoration effect at the edges and corners, and the smooth effect is suppressed.Secondly, a joint blind deblurring method is improved to solve the edge diffusion problem occurred in the process of recovery. It is a kind of Tikhonov and TV mixed constraints method. First of all, a edge detection filter is used to decompose the image into the edge region and the smooth region. Then, the edge region is constructed with TV constraint, and the smooth region is constructed with the Tikhonov constraint. By using this model, the edge information is prevented from spreading to the smooth region,which can solve the problem of edge diffusion. Aiming at the problem of the amplification effect of small positive noise in the image and PSF iterative process, asmall positive threshold constraint split Bregman method is used for numerical solution.This method improves the non negative constraint condition of the traditional method to a small positive threshold. At the initial stage of the iteration, the small positive noise is suppressed, and then the PSF and restored images are obtained. The experimental results show that the proposed method is applicable to all kinds of fuzzy types and has better performance in the real image restoration.To sum up, the thesis studies the TV blind deblurring method from two aspects,and improves the effect of image restoration, which lays a solid foundation for the practical application of image blind deconvolution technology.