Research On Blind Recovery Of Noise Blurry Image Based On Filtering

Image blur is a common phenomenon,which is usually caused by camera shake,inaccurate focus,object movement in the scene and other factors.Image blur not only greatly reduces the image quality,but also has a great impact on the subsequent image application.The main purpose of image deblurring is to recover clear image from blurring observation data and to restore its original details and features as accurately as possible.It consists of two sub-problems:blur kernel estimation and image deconvolution.The accuracy of blur kernel is very important for image deblurring.In recent years,image blind deblurring is a hot issue in image processing.Blind image deblurring refers to the task of restoring clear images from blurred images without any prior information about the image or blurring process.This is an ill-posed inverse problem,knowing neither the underlying sharp image nor the blur kernel that causes the blurring.In order to solve this problem reasonably,a priori information is necessary.At present,many technologies and algorithms have been used to solve the problem of blind image deblurring,but in the field of image processing,blind image deblurring is still a challenging task.The main challenge arises from blur kernels that are difficult to estimate accurately,especially when observation images lack useful kernels estimation information.At the same time,we find that many blind deblurring methods assume that the blurred image is noiseless.Therefore,when the image is blurred and accompanied by strong noise,the processing effect of many algorithms is poor.In order to solve this problem,a new blind restoration method based on image filtering for noise-blurred images is proposed.The model in this paper is based on the optimization framework of maximum a posteriori estimation(MAP),which can process both normal noisy blur images and noisy blur images with outliers.The data fidelity item can process outliers and normal pixels adaptively.During the solution,the iterative reweighted least squares(IRLS)algorithm is used to minimize the problem as a series of standard least squares problems,each of which is reweighted by the solution of the previous step.In order to accelerate potential image estimation,we introduce a prediction step in the iterative deblurring process.In noise suppression,we introduce guided filter.Guided filter has the property of preserving edge and suppressing noise.The introduction of this method can suppress the noise of latent image well in the process of iteration and reduce the influence of noise on kernel estimation.A large number of experiments show that the method in this paper achieves good results on many images,and realizes the blind deblurring of noise-blurred images well.