Motion Debluring Using Spectral Properties And Image Priors

Image deblurring is a fundamental but challenging problem in computer-vision research.The formation of a blurred image is caused by a variety of things: camera shake,object motion,camera out of focus,etc.,which also makes the deblurring work complicated because it needs to consider many factors.Image deblurring is mainly divided into non-blind deblurring and blind deblurring according to the known or not of blur kernels,while blind deblurring is a kind of work with less known information and higher difficulty in the deblurring work.In the process of blind-deblurring,the prior work of blur kernels and latent images is very important.Based on the following observations: 1.The dark channel of the clear image is more sparse than the blurred one;2.The frequency of the clear image in the frequency domain is significantly higher than that of the blurred image.On the basis of the existing deblurring method,we apply the sparse coding and dark channel theory respectively to the clear image prior,constitute two effective image priors,combined with the effective blur kernel prior which based on the convolution spectrum to deal with the deblurring.The main research results are as follows: 1.Two kinds of image deblurring algorithms combining sparse coding,dark channel theory respectively used in image prior,and the deblurring results are surprising.2.Under the guarantee of deblurring effect,the speed of algorithm iteration convergence also has a certain extent increases.3.With the two works using from the sparse coding to the dark channel theory,have progress on my self.