Intensity based and edge based segmentation methods have their own limitations. Multiscale representation employing Gradient dependent diffusion and wavelet decomposition, based on edge and intensity information, has been proposed in the literature to get rid of these limitations to some extent. In this work, a fully automatic segmentation method based on complex diffusion is proposed. A multiscale representation of the image is formed based on the nonlinear complex diffusion technique. Intensity based linking model is used to group pixels into a number of segments. This approach is employed successfully to segment MR Brain images into White Matter (WM), Gray Matter (GM), and Cerebral Spinal Fluid (CSF) with no user interaction. Better segmentation performances were observed when compared to Gradient dependent diffusion and a'trous based methods proposed in the literature.
Jyothisha J. Nair and Govindan, V. K., “Multi-Scale Segmentation Based on Nonlinear Complex Diffusion”, Journal of Medical Imaging and Health Informatics, vol. 3, pp. 242–245, 2013.