Research In Restoration Technology Of Blur Images Based On Motion Model

This paper studies blurred image restoration problem caused by the relative motion between the subject and the camera movement. By analyzing the motion imaging principle of blurred image, image blurring caused by movement can be approximated as uniform linear motion in the short exposure time of the pickup image. This paper builds image degradation model and recovery model. In the recovery process, how to establish the point spread function accurately is the key issue to affect the quality of image restoration. In this paper, we identify motion blur parameters based on the autocorrelation and wavelet transform. After motion blur point spread function has been estimated, we restore the blurred image using recovery methods in order to obtain the restored image which is similar to original image and eliminate the ringing effect.In existing recovery algorithms, inverse filtering performs well when degraded images do not contain noise; but when there is noise, we can hardly see the original image from the restored image. Wiener filtering has certain antinoise ability, but its ringing effect is obvious. RL recovery algorithm is time-consuming. If the image contains no noise, the more iteration times, the closer to the original image, but when image contains noise, each time of iteration will enlarge the noise, and make the recovery effect variation.This article researches recovery theory focusing on motion model estimation, with wiener filtering algorithm as the main recovery tool, combined with image to ringing effect a full image restoration theory. Including:by analyzing the motion model, and the model of uniform linear motion imaging model; motion blurred image PSF estimation, this paper introduces the method of frequency domain estimation parameters, PSF estimation method based on the wavelet transform, Hough transform and autocorrelation function; the existing image restoration algorithm research, including inverse filtering method, the wiener filtering method and RL iterative method; removing ringing effect method research.