Study On Blind Image Restoration Algorithms

Image restoration includes the Inverse Filter, Wiener Filter, Constrained Least Squares Filter and other classical algorithms. Blind Image Restoration is a focus in image restoration field. In this thesis two methods of BIR --Priori Blur Identification and Iterative Blind Deconvolution -- are illuminated in detail. Blind Image Restoration is the process of estimating both the true image and the blur from the degraded image characteristic using partial information about the imaging system. Priori Blur Identification gets the PSF before restoration implementation, while Iterative Blind Deconvolution estimates the true image and the PSF at the same time.A new method to estimate the parameters d,θ, R of Point Spread Function and k of Wiener Filter is brought forward. Because the degraded image includes the position information about frequency domain zeros of PSF and that position is related to the parameters of PSF, blur type and parameters can be determined exactly by using these conditions. With the use of E-k error curve, the optimal value of k can be chosen accurately. Then the optimal result of restoration is obtained. This parameter estimation method is self-acting, precise, facile and fast. It is effect to the motion blur of any direction, defocus blur or the mixed blur. Lots of experiments have showed its validity.On the other hand, some research has been made on the Iterative Blind Restoration. In the latter of this paper, the IBD algorithm is interpreted and realized, which is helpful to the image with finite bound and uniform background. Several ideas and future work are written in the end of the paper based on the conclusion of forenamed methods.