A new fast segmentation algorithm (for colorful images) based on characteristic of dispersion of RGB

A new image segmentation algorithm is presented based on characteristic of dispersion of RGB(Red, Green, Blue), utilizing the information of the colorful image to achieve fast segmentation. With tow images an example, firstly, in the original image, it gets the sum of all pixels' RGB dispersion, then that is normalized into grads image. In the histogram of RGB dispersion the characteristic is Bi-peak. Based on this, using one iterative algorithm automatically selects the optimal segmentation threshold. Experimental results indicate that this algorithm can segment colorful image accurately and fast, and adaptable for checking of high-speed moving object, which is of great value.