A Fast Deblurring Algorithm for Surveillance Scene

This paper presents a new resume algorithm for video image which is disturbed by motion blur. For innovation, the new algorithm makes the denoised region of motion object as the primary target, and with the discrete model, it gets valuable result. Firstly, the method calculates the differences between the neighboring frames. Secondly, it calculates the quadrinomial moment pixel by pixel to differentiate the motion object from the background noise, with the results, it gets the region of object and abstracts the motion vector. And finally, only aiming at the object region, it uses the motion vector to accomplish the deblurring processing with the discrete model. Extensive experiments show that the new algorithm has fast processing speed, and it can reduce the disturbance of background noise, so the new algorithm improves the deblurring efficiency.