Research Of Shake Compensation Techniques In Image Imaging Process

Analysis of the process of camera shake, according to the length of exposure time, camera shake blur image will be divided into two types, uniform motion blur distortion image and double shadow blur distortion image.Shake blur image will be gone through the two-dimensional Fourier transform. This transformtion values will be processed through normalization and binarization. Combining with the actual situation of blur images,shake blur direction can be estimated through the algorithm of maximum based on Radon transformation (MRT). Rotating shake blur image to the horizontal axis and performing one-dimensional derivative in the horizontal direction,combining with the actual situation of blur image,shake blur length can be estimated through the algorithm of pitch of derivative autocorrelation (PDA).The point spread function (PSF) of uniform motion blur distortion image looks like a rectangle. In MATLAB, by directly call the created function of point spread function and the Wiener filter image restoration function, image restoration can be realized. The point spread function of double shadow blur distortion image is more complicated. In this dissertation, three approximate models are presented, including slope type, sinusoidal type and parabolic type. In MATLAB platform, first create the corresponding point spread function, then call the Wiener filter image restoration function, and finally image restoration can be realized.When the above methods applied to a variety of actual shake blur image, the image restoration works well.