Study On The Restoration Methods For Blurred Images

Mage restoration techniques have been widely applied to many areas such as astronautics, navigation, artificial intelligence, control theory and people's entertainment. During the process of taking digital images, we often obtained the blurred ones due to some uncontrolled ingredients such as the accuracy of the system, median of transport, atmosphere circumstances and wobble of the machine, which directly result in bad effects on later applications. Therefore, it is important to restore the original images from the blurred ones as promptly as possible. In this thesis, we mainly study the image restoration techniques for the motion blurred, the defocused blurred and the weather blurred pictures. We first introduce the basic techniques of the image restorations. For the images degraded from the separable PSF, we then propose a conjugate gradient method for solving a Sylvester-type equation based on the least square principle. We also construct two forms of preconditioned methods and discuss their implementations. To decrease the computational complexity of image restoration, we devise a new deblurring algorithm based on a new-developed anti-reflective boundary conditions. This method divides the original image into several sub-blocks and incorporates the anti-reflective boundary conditions, which overcome the discontinuity of the normal derivative brought about by other boundary conditions. Numerical experiments show that the new developed algorithm, by using proper sub-blocks and width of the bound, outperforms the original one with respect to the time used.