Novel diffusion algorithm based on local statistical parameter for image de-noising.

Aiming at the problem that isotropic algorithm can easily blur the feature of image information and that tensor-based algorithm will produce pseudo striations at noisy points within homogeneous regions, this paper proposed a feature parameter that was local variance of image, this parameter could effectively reflect local structural features of image, based on this parameter, it proposed a novel algorithm that combined the above two algorithms. The proposed algorithm could adaptively adjust to the diffusion rate of the above two diffusion algorithms according to local statistics of image when de-noising on different regions of the image in order to preserve the advantages and cancel the drawbacks existed in the two algorithms mentioned above. The experimental results show that the proposed algorithm compared with others classical algorithm can remove the noise effectively and preserve, enhance the image edges better. [ABSTRACT FROM AUTHOR]