Research On Iterative Multi - Channel Image Blind Restoration Algorithm

Image restoration has the characteristic of practicability and application of universality, which make it get long-term development. The process of image restoration is to construct the mathematical model of image degradation process firstly and solve it accordingly. According to different solving process, the image restoration can be divided into non-blind restoration and blind restoration. Due to the practical application of blind restoration algorithm, it gets gradually developed. Based on image restoration technology, scholars have launched a blind restoration technology research. Now, there are many algorithms, such as the model based on total variation, or based on minimum L1-norm, or based on Bayesian prior, or based on the high-order noise, or based on markov random field and so on. Among these algorithms, single channel blind restoration takes shorter time, but the effect is not very good. Multi-channel blind restoration can get better result, but it will take longer time. To solve these problems, this paper conducted the following research:(1) In the basis of analyzing the existing algorithm, I conducted some experiments to verify the classic image restoration methods and analyzed its recovery effect. Then I analyze the feasibility of multi-channel blind restoration algorithm’s, by changing the noise standard deviation, to verify the original algorithm’s robustness to noise. By changing the size of the point spread function (PSF) to analyze the influence on the effect of image restoration PSF. By doing experimental comparison to analyze the advantages and disadvantages of multi-channel blind restoration algorithm with classical reconstruction method, and to make a conclusion about the multichannel blind’s superiority to single channel re-convolution.(2) There are two problems in the existing multi-channel blind restoration algorithm, one is that it takes longer time; another is the inflexible of initial point spread function (PSF). To solve these problems, I combine the minimize the multi-channel alternating iterative algorithm with radon transformation and Fergus recovery algorithm, respectively. Firstly, I use the radon transform and Fergus recovery algorithm to do initial recovery on blurry image, respectively, and then to estimate the point spread function (PSF) as a multi-channel blind restoration initialization point spread function for blind image restoration. Compared with the original method, multi-channel blind restoration algorithm takes shorter time, has higher signal-to-noise ratio, the effect of recovery is better too.