A multistage image segmentation and denoising method - based on the mumford and shah variational approach

Abstract

A new multistage segmentation and smoothing method based on the active contour model and the level set numerical techniques is presented in this paper. Instead of simultaneous segmentation and smoothing as in [10], [11], the proposed method separates the segmentation and smoothing processes. We use the piecewise constant approximation for segmentation and the diffusion equation for denoising, therefore the new method speeds up the segmentation process significantly, and it can remove noise and protect edges for images with very large amount of noise. The effects of the model parameter v are also systematically studied in this paper. © Springer-Verlag 2004.