| In this thesis the basic principle of maximum entropy method (MEM) and its applications in image restoration are thoroughly discussed. Programming with MATLAB, most exist algorithms of maximum entropy restoration are tested by various artificial degraded images and all achieved better performance than classical Wiener filtering. Bonavito's algorithm is verified to be specially suitable to restore complicated images in a normal PC because of its good performance and short processing time, and so is Zhuang's algorithm which is extended here for coping with the cases of low noise level and sparse data. Methods of PSF and noise variance estimation are proposed for practical restoration and thanks to them a real defocus image is perfectly restored. |
PDF Full Text Request