Design method for nonlinear diffusion filter and its application

Nonlinear filters show good performance in preserving important features. These filters may boil down to solving nonlinear diffusion equations whose initial values are input images. Designing appropriate diffusion coefficients can control the action of diffusion equations, so the important features of images can be preserved while the images are smoothing. A design method for nonlinear diffusion filter is proposed. Choosing an appropriate coordinate by introducing image feature, we turn the design problem of diffusion matrix into the design of diffusion coefficients along local coordinates by a coordinate transform. This design method for the diffusion filter is simple, and helps us to understand the diffusion action of the diffusion filter and has obvious physical meaning. Some experimental results are given to show that this design method is feasible, and reaches an obvious effect in terms of edge-preserving image processing.