Complex Diffusion Based on Topological Derivative for Image Denoising and Edge Extracting

Zhang Wenjuan,Feng Xiangchu

DOI: https://doi.org/10.3969/j.issn.1004-9037.2010.03.012

2010-01-01

Abstract:An algorithm concerning application of nonlinear complex diffusion is presented based on topological optimization to image denoising and edge extracting.The linear diffusion blur edges,based on the idea of topological optimization,linear complex diffusion coefficients at each pixel are perturbed.It makes the lowest topological derivative to be optimum one.Diffusion coefficients have anisotropic property,thus overcoming the flaw of isotropic diffusion that cannot remove the noise along edges.Then,pixels with enough small topological derivatives are chosen,and the diffusion is applied by using optimum diffusion coefficients.The stop criteria of the algorithm are introduced.Experiments show that compared with the nonlinear complex diffusion by Guy Gilboa, the real image of the processed image by the algorithm show better denoising effect.Additionally,the edge is well preserved on the imaginary image.The algorithm can remove the staircase effect produced by Perona-Malik method.