Research Of Image Restoration Based On Prior Model With Local Constraints

Digital image restoration has widely concerned by the academic community in the field of image processing.Nowadays,people are getting higher and higher in the quality of images.Therefore,it is great significance to reserach for image restoration algorithm with high-performance.Image restoration is an ill-posed inverse problem,which can be solved by regularization technique.The study for regularization parameter selection is an open problem in image restoration.Regularization parameter can improve the performance of image restoration for its balance of fine degree of peace and smoothness.In order to preserve more details of images,we combine the image prior model with local constraints to construct a new image restoration method.In this paper,we construct a novel image restoration method with adaptive regularization parameters.The main work is as follows:1)By imposing local constraints,we propose an adaptive estimation method of regularization parameter for expected patch log likelihood(EPLL)based image denoising,with the consideration that the Gaussian mixture model has the capability of clustering.Expected patch log likelihood based image denoising approach has been shown to be surprisingly competitive in image restoration.However,recent related works generally utilize global regularization parameter that influences the performance of denoising algorithm.By imposed local constraints,our method jointly employs the Lagrange multiplier technique and local entropy concept to select regularization parameter for each underlying cluster.Experimental results illustrate the relatively good performance of our image denoising method in terms of visual improvement and peak signal to noise ratio.2)We introduce a spectral response(SR)based total variation image restoration method with adaptive regularization parameter by construct local constraints.Proper regularization parameters can not only improve denoising and deblurring,but also maintain excellent image detail.This method uses the SR to construct local constraints,for adjusting the weight of the prior item and the fidelity term of the total variation model.Moreover,in order to fast calculate the model solution,we employ the alternating minimization numerical technique.Experimental results show that a satisfactory result can be obtained by this method at lower time consumption.