The Research On The Parameter Identification And Restoration Of The Rotation Motion Blurred Image

In the production or military equipment, a growing number of image devices have been installed on the rotating platform, inevitably, imaging equipment which working on the rotary motion platform will encounter the restoration of rotating motion blurred image, so the study of rotating motion blurred image restoration has very important value of practical application.In the thesis, we study the rotary motion blurred image. On the basis of the linear blurred image, the rotary motion blurred mathematical model is deduced, and then obtaining the point spread function(PSF) in the form of matrix. At the same time, this thesis introduces the three kinds of linear motion blurred image parameter estimation methods: Fourier transform method、Error-parameter analysis method and Differential auto-correlation method. Meanwhile, we analyzed the effects of these three methods.this thesis presents a method based on the differential auto-correlation for the parameter estimation of rotary motion blurred image. The method detects the fuzzy center and fuzzy angle relying on the linear relationship between the blur scale and radius on the fuzzy path. Firstly, we determine the scope of rotation center on rotary motion blurred image by Canny edge detection and Hough transform method.Secondly, we select each pixel as the center of rotation in certain range, and then extract the pixel points in different radius by Bresenham method, and transform them into linear form, and then calculate blur scale of different radius by differential auto-correlation. Finally, we can obtain the the fuzzy center and fuzzy angle by observing the fitting line of scatter plot which shows the relationship of blur scale of each fuzzy path with the corresponding radius in coordinate space. The experiment shows, it’s effective that the method is introduced to identify the parameter of the rotary motion blurred image.The identification of rotary motion blurred parameter is the key to determine the image degradation function. However, most of the degradation functions have singularity, so the image restoration problem is ill-posed problem. The regularization method is the common way to solve ill-posed problems. This thesis introduces the image sparse representation method and studies the structure of the sparse dictionary.This thesis lays out mixed norm regularization method basing on image sparse representation to solve the problem of rotary motion blurred image restoration. This thesis selects a efficient algorithm as separable surrogate function method(SSF) tosolve the hybrid model basing on theoretical analysis and wavelet dictionary constructed by using prior knowledge of image. We convert the rotary motion blurred image into the linear motion blurred form, and then solve the sparse vector of pixels which come from different fuzzy paths by SSF algorithm. we obtain new vectors by combining the spares vectors with wavelet dictionary, and get the clear image through inversely transforming new vectors. The experiment shows that it’s effective that the method proposed in this thesis.