Color image edge detection and classification based on physical model

A new color image edge detection and classification method based on a physical model (dichromatic reflection model) is presented in this paper. First, we analyze this model and discuss color invariant features in different color spaces. Then, a new color image Sobel edge detector with adaptive threshold based on class variance is proposed. With this method, color image edge detection is carried out in different color spaces. At last, edges are classified into one of the following classes according to different color invariant features: material edges, shadow edges and highlight edges. Experiments show that this edge detection and classification method can get nice result and has good application prospect.