Noise removal using nonlinear anisotropie diffusion filtering based on statistic-local open system

In this paper a novel nonlinear anisotropic diffusion filter model based on statistic-local open system is proposed. In impulse noise removal using the new model, only estimated noised pixels are processed, unnecessary blurring caused by pure pixels energy diffusion can be reduced. In the new filter model, some pixels are devised as "origin" pixels and "convergence" pixels to make up energy loss and eradicate noise energy. And a newly designed conduction coefficient is adopted to avoid energy flow from noised neighbor pixels. We test the performance in unipolar and bipolar impulse noise removal respectively and compare filtering effect using P-M, Catte, Median and proposed filter. Experimental results demonstrate the improvement in noise removal and edge preservation.