A Method Of Image Blind Motion Deblurring

With the development of science and technology,mobile phones,digital cameras and other tools allow people to easily capture the moments of their lives.However,due to the photographic environment and the photographic process,pictures taken are prone to blur.Due to the diversity of the causes of deblurred images,there is currently no market-oriented application.However,in the military and criminal investigation fields,it is of great practical significance to be able to restore deblurred images clearly.The value it brings is inestimable.This makes the study of image deblurring a hot issue in the field of image processing.In recent decades,research on image deblurring has produced new results.In particular,the proposal of the dark channel as an image priori has been proposed,and the problem of image deblurring has been solved more effectively,but this method does not apply to noisy images.Once the image contains significant noise,the resulting deblurred kernel is discontinuous,and the desired clear image cannot be recovered.This paper takes the motion blur image as the research object.By using the Radon transform technique,Radon transform of the filtered kernel corresponding to the filtered image and the Radon transform of the original image corresponding to the deblurred kernel are obtained firstly,and the equality is used as the starting point to improve the multi-scale darkness.The prior method of the channel makes it suitable for deblurred images with obvious noise,and thus makes up for the insufficiency of the dark channel method in noise processing.This paper follows the treatment of nonlinear dark channel operators in the dark channel prior method,uses Radon transform techniques to update the deblurred kernel,and adds a non-blind deconvolution method that effectively suppresses noise.In this paper,experiments on database images and a large number of real images are performed.First,the importance of using Radon transform techniques for deblurred kernel estimation is demonstrated.Second,the number of direction filters is analyzed by comparing the effects of the number of different direction filters on the experimental results.The shape similarity of deblurred kernels is directly proportional.Third,by comparing the effects of different levels of noise on deblurred kernels,it is shown that this method can handle blurred images with noise.The experimental results show that this method can make the blurred natural image and face image clear.