Research On Blind Restoration Of Motion Blur Images

The degradation of image quality is an inevitable problem in the application of digital image.Many factors could cause degradation of image quality,such as atmospheric turbulence,defocus,and relative motion between target and camera.Motion blur is the most common situation in the degradation of image quality.In order to improve the image quality,many deblur algorithms have been proposed,which proved to be effective in deblurring images,sharping image edges and extracting useful information.However,most of the existing deblur algorithm still have limitations: some algorithms could not obtain a high-grade result,while others are too complicated to apply.These limitations impose restrictions to solve practical problems.This paper studied the progress of image degradation and analyzed the advantages and disadvantages of existing deblur algorithms.Then a single image motion deblur algorithm was proposed and extended to the multi-image situation.Specific works are as follows:First of all,after analyzing the basic theory of image degradation,this paper proposes a single image blind motion deblur algorithm.The algorithm uses image preprocessing technology to predict the derivatives of latent images from blur images.Then the predicted result is used to estimate the Point Spread Function(PSF).At last,the latent image is estimated via deconvolution method with the help of PSF.The simulation and experiment results show that the proposed algorithm could restore the latent image effectively.The image quality could be enhanced significantly.Secondly,after studying the property of multi-image motion deblur problems,a multi-image blind motion deblur algorithm was proposed.The advantages and disadvantages of multi-image motion deblur problems is analyzed at first compared to the single image situation.In this algorithm,the result of single image blind motion deblur algorithm is regarded as reference in order to register blur images.Subsequently,latent image and PSF were estimated alternatively via Maximum A Posteriori model.Results proved that multi-image motion deblur algorithm performs better on deblurring motion blur images.The main research of this paper is about blind motion deblurring.Our work is a kind of enhancement of the existing methods.The algorithm proposed in this paper could keep the details of images while restraining the ringing artifacts.The work of this paper makes a foundation of follow-up tasks and has an important significance on motion deblurring.