The Application Research Of Video Motion Deblurring Technology

Whether it is a single image or video sequence, restoration of degraded images is quite realistic research and has considerable prospects, however, there are still many thorny problems to be solved. Motion blur is a very common phenomenon of image degradation. Our study is major in video motion deblur, when the background is static but the foreground move fast within the exposure time of the imaging system causes blurred image, and we focus on how to deblur this kind of motion blur.Firstly, this paper presents some related algorithms of initial research and some experimental results and analysis, then propose our final video motion deblur algorithm framework. Therefore, this paper is divided into five major modules:(1) The initial extraction of foreground algorithms, those algorithms include background restoration algorithm based on pixel intensity classification, robust matting, Trimap automatic generation based on random walker. And then elaborate some defect and some idea get from experimental results.(2) The initial estimation of PSF algorithms, 1-D and 2-D motion deblur using transparency information. And then elaborate some defect and some idea get from experimental results.(3) Extraction of blur foreground of video frame, using a closed-form solution to natural image matting, then the mark is automatically passed to adjacent frames, this algorithm is introduced to the process of matting, so we can successively mark video frames.(4) Estimation of PSF(Point Spread Function), this paper proposes an algorithm: using SIFT matching blur foreground of adjacent two frames, get the motion vector in inter-frame time, weighting the proportional relation between inter-frame time and exposure time to the motion vector in inter-frame time to get the motion vector of exposure time as the initial motion vector, then feedback the restored clear binary map transparency to estimation of PSF to get the excellent PSF.(5) Image restoration, Image Fusion Technology added to the iterative process of R-L deconvolution algorithm, the iterative convergence to clear foreground image with effectively suppressed ringing.Comparing the experimental results shown we ultimately obtain a relatively shorter time-consuming, relatively better deblurring results and more practical algorithm.