Color image segmentation with an entropy-based cost function

We present a novel color image segmentation algorithm which incorporates an entropy measure with the spatial information of the image. Using entropy, the algorithm creates a cost function which is used with the histogram-based probability distribution function of the color components of the image. Segmentation is achieved through dynamic programming that optimally partitions the histogram of each color component. An output image can have any number of color levels from two all the way up to the original number of colors present. The reported simulations of the algorithm on color images in the RGB and HSV color spaces give very good results compared to many existing methods, while maintaining low computational complexity in terms of storage and processing requirements.