Improved method of parameters identification and restoration of motion blurred image

Abstract

In this paper, dealing with the original image as an isotropy one rank Markov process, we use present weighted average method based on directional derivation to identify the direction of motion blurred image. The identification results show that the direction of motion blurred is not only influenced by the direction of motion blurred, but also influenced by the object shape. So we give a new way to identify the motion blur direction from the blurred image by improved direction derivation method. Under the consideration that one object can't occupy four corners of the whole image, the new idea of identifying four corners of the image instead of the whole picture is proposed. It can identify any direction, from -90° to 90°, with high precision and high stabilization. The experimental results show that the motion blurred direction can be identified effectively by the new method. The mean square error is reduced to 68.55% compared with the old method. The blurred image can be rotated to a horizontal axis according to the motion blurred direction. For the pixels in blurred images have high correlation with the neighbors, we use derivation and correlation methods to estimate the blur parameters. At last we complete the restoration of motion blurred image by Wiener filters.