Image Restoration From Blurred Object Of Linear Motion

In recent years, with fast development in the digital photography market, themonitoring system has been widely used in car speeding detection system, air defensesurveillance system, home surveillance system, and license plate recognition system.However, photographic imaging will be influenced by air disturbance, machinevibration, motion speed, equipment settings, etc., and hence the image captured bycamera will be blurred which is hard to recognize. This motivates that how to derive aclear image owing to the motion-blurring problem becomes a hot research topic. Basictheories are introduced on motion blur image, such as properties of Fourier Transformabout the image, and image degeneration model. Three restoration principles ofinverse-filtering, wiener filtering, and Richardson-Lucy are explained and compared. Inthis thesis, a real-time restoration method for linear local motion-blur image is proposed.The proposed method is aimed at restoring the local-blur image, unlike most reportedmethods for restoring the whole blur image. For an image in which only the fastmoving-object is blurred but the background is clear, the proposed strategy is to firstlydivide such an image into many subimages and then detect the blurred subimage by thegradient distribution and the maximum of cepstrum. For a blurred subimage, the blurdirection and blur length are estimated to calculate the parameters of point spreadfunction (PSF). Richardson-Lucy deconvolution algorithm is employed to restore thisblurred subimage. Using many artificial and real blurred images with various contents,the experimental results show that the proposed approach is more accurate and robustthan other methods.