Blind Image Deblurring Via Prior Preserving

With the wide application of computer vision in various fields such as scientific research,communications,medical care,and security,the requirements for image quality and clarity have become higher and higher.Image blind deblurring has also become a hot issue in the field of computer vision.The key to blind deblurring of images lies in the blur kernel estimation.At present,the information of grey blurred image is mainly used as the input of the blind deblurring model to estimate the blur kernel,and than is used to deconvolute clear images with the color blurred image.Therefore,whether the gray blurred image can maintain the information of the original color blurred image to the greatest extent directly affects the estimation of the blur kernel and the recovery effect of the clear image.As the current image blind deblurring method does not consider the problem of information retention in the process of image graying,we propose a blind deblurring method based on prior preservation,considering the prior knowledge of blur images.The main work and contributions of this article are:1.Aiming at the phenomenon that the gradient sparsity of the color blurring process in the existing deblurring method can't be well maintained,we propose an image blind deblurring method based on contrast priors.The method uses a grayscale method based on contrast preservation to process the color blur map,so that the grayscale blurred image retains the original color map gradient information to the maximum extent.Using the improved grayscale blurred image as the input to the deblurring model improves the constraints on the similarity of the model and the sparsity of the image gradient,and is more conducive to the estimation of the blur kernel and the improvement of the deblurring effect.A large number of qualitative and quantitative experimental results show that this method can effectively improve the deblurring effect and is superior to existing algorithms.2.In order to maintain the dark channel information while preserving the contrast of the original color blurred image,the blind deblurring method for contrast/dark channel priors is proposed.On the basis of the contrast-preserving gray-scale algorithm,the dark channel retention constraint is added,and the loss of dark channel prior information in the gray-scale process is reduced.Through extensive analysis,it is proved that this method is more constrained to gradient sparsity and dark channel sparsity of clear images than method one,thereby further improving the estimation of the blur kernels and improving the effect of deblurring.The experimental results show that the effectiveness of this method for blind deblurring of images is proved by both subjective and objective evaluation,and the deblurring effect is better than those of method one and other methods in the field.The blind deblurring method proposed in this paper around the prior preservation of the blurred image overcomes the problem that the information of the grayscale image in the color blurred image cannot be well maintained,and strengthens the regularization constraint of the deblurring model.This experiment uses the image deblurring kohler data set,and we compare the results with a number of current leading deblurring algorithms.The experiments show that the proposed method has good performance and the deblurring effect is better than other methods.