Image blind deblurring method based on external prior information of image block and sparse representation

Abstract

The invention discloses an image blind deblurring method based on the prior information of an external image block and sparse representation, and specifically relates to the external prior estimation of an image block and the application of the external prior estimation into an image deblurring frame. The method comprises the steps: carrying out the training of an image in a clear image library, obtaining a Gaussian mixed model, describing the external prior knowledge of a fuzzy image block through image block expectation logarithmic likelihood estimation, enabling the external prior knowledge to serve as a regular term and to be added to the deblurring frame based on the sparsity, carrying out the iterative restoration of a central image and the solving of a fuzzy kernel in the frame based on the sparse deblurring; constructing a dictionary of a single central image block through a single Gaussian covariance matrix of each type, obtaining a sparsity coefficient through the dictionary, and constructing a central clear image; solving the fuzzy kernel through an augmented Lagrange algorithm; finally solving a final clear image through a hyper-Laplacian algorithm in non-blind deconvolution. An experiment result indicates that the method is better in effect of ringing inhibition and noise weakening.

Description

technical field [0001] The invention relates to a method for image deblurring, in particular to an image blind deblurring method based on a priori information of external image blocks and sparse representation. Background technique [0002] Due to the relative motion between the camera and the shooting scene, the obtained images often have a certain degree of motion blur. The model of image degradation can be expressed as the following convolution process: [0003] [0004] where Y is the obtained blurred image, K is the blur kernel, X is the clear image, and N is the noise. [0005] Due to the important application value of image deblurring, it has received extensive attention at present. More and more prior knowledge of natural images is applied to the deblurring framework, literature (R.Fergus,B.Singh,A.Hertzmann,S.T.Roweis,andW.T.Freeman,Removing camera shake from a single photograph,ACM Transactionson Graphics, vol. 25, no. 3, pp. 787–794, 2006) proposed that on t...

AI Technical Summary

Problems solved by technology

The above-mentioned methods are all based on deblurring the overall image, and do not take advantage of the characteristics between image blocks

Images