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.
Noise Removal Using Nonlinear Anisotropic Diffusion Filtering Based on Statistic-Local Open System
2008 Congress on Image and Signal Processing ; 3 ; 372-378
2008-05-01
1345225 byte
Conference paper
Electronic Resource
English
Image noise removal approach based on subpixel anisotropic diffusion
British Library Online Contents | 2012
|Robust Lee local statistic filter for removal of mixed multiplicative and impulse noise [5298-32]
British Library Conference Proceedings | 2004
|An Anisotropic Fourth-Order Diffusion Filter for Image Noise Removal
British Library Online Contents | 2011
|Two-stage radar image despeckling based on local statistic Lee and sigma filtering [4304-16]
British Library Conference Proceedings | 2001
|Nonlinear vector filtering for impulsive noise removal from color images
British Library Online Contents | 2007
|