Abstract:

A nonlinear partial differential equation model was proposed to remove multiplicative noise of  color images. The geometrical manifold characteristics of vectorvalued image after smoothing are introduced to  distinguish the “real” and “false” edges. The p(x)-Laplace diffusion coefficients drive image to diffuse in different ways on the edges and in the domains of images. The |u|^q diffusion coefficient drives image to diffuse at different speeds that fit to the multiplicative noises strength. Theoretical analysis and numerical results show the efficiency of the new method.

Key words: color image, multiplicative noise, image restoration, anisotropic diffusion