Anisotropic Diffusion and Segmentation of Colored Flowers

This paper presents a new anisotropic diffusion method for multi-valued images. The colored flowers are segmented from the background. In my diffusion technique the edge stopping function is computed as inversely proportional to the weighted sum of gradients of color components. The value of edge stopping function tends to zero as the weighted sum of three color gradients becomes large. The other algorithm further explores an existing diffusion technique based on structure tensors. The multivalued image diffuses using a system of coupled differential equations in the direction of minimal change. The paper presents the results of anisotropic diffusion and segmentation of colored flowers. The new method can also be applied for anisotropic diffusion of other multivalued images.