Research On Motion Blurred Image Restoration

As the relative motion between object and camera in the process of image acquisition, theacquired image is a certain degree of fuzzy. Motion blur degrades quality of the image, whichseriously influence on people seeking to the outside world, access to information and thedevelopment of scientific research. In order to solve the problem of motion blurred imagerestoration, this paper studies on image restoration algorithm and blurred kernel estimationtechniques in the unknown blurred kernel，which proposes a new theoretical method onprediction technique of blurred kernel, and establishes the framework of the motion blurredimage restoration algorithm. The main work is divided into the several aspects:Firstly, it researches the basic theory of blurred kernel prediction technology. As atraditional forecasting method of blurred kernel, the theory of extracting blurred parameters byusing spectrum distribution of fuzzy image seems unlikely to work and exists alarge error disadvantages. To solve this problem, the paper studies on characteristics of naturalimage and the causes of blurred image. Natural image has clear edge image, and the edgeinformation can easily be extracted. Taking this as the breakthrough point, this paper usesshock filter to enhance edge of blurred image, which could predicts sharp edge from theblurred image. And it will estimate blurred kernel combining sharp edge with blurred imageedge information. This method can reduce error and estimate more ideal blurred kernel thantraditional prediction methods.Secondly, it has deeply studied on the image gradient distribution model of the restorationalgorithm. It focuses on analyzing the characteristics of the two kinds of models that is Gaussdistribution and hyper-Laplace distribution image restoration algorithm. Gauss distributionmodel restoration algorithm is adopted to estimate the fuzzy kernel function, and it useshyper-Laplace distribution model to rebuilt blurred image based on estimating accurate blurredkernel. At last, this paper designs the framework of blurred image restoration algorithm andprovides the main flow chart.From last test results, the algorithm in the paper has good estimation effect and highworking efficiency of the blur kernel estimation, which has a good robustness against noiseusing the salient edges. Moreover, it can get high quality estimated image in the subsequent blurred image restoration work. Compared with the restoration effect of traditional algorithm,it has better restoration effect and gets high quality estimated image in this paper.