Unsupervised Segmentation for Color Image Based on Graph Theory

Image segmentation method based on graph theory is mainly used for gray images, and thresholding of segmentation should be predefined. Combining with entropy in information theory, this paper suggests an unsupervised method for color image segmentation. The image is mapped into an weighted undirected graph, the pixels are considered as nodes, the best thresholding is obtained by objective function of maximum weighted entropy to realize unsupervised segmentation. Experiment results show that the new algorithm ensures the color image segmentation excellent disturbance attenuation performance and better separability.