Noise reduction based on reaction-diffusion system

Noise reduction is realized by using a FHN (FitzHugh Nagumo) model which is a typical two-variable reaction-diffusion equation. The model exhibits two characteristics of dynamics: excitability and bistability. In the bistable region the local dynamics of the system can leads to the availability of noise reduction. It is found that the effects of image smoothing by FHN model are superior that by other filters. Results show that this approach can not only remove the noise effectively, but also have the advantages of the continuity, integrity. Furthermore, accurate location of image edges can be preserved during the smoothing process. The mechanism of the new method has been studied by the linear stability analysis.