Abstract: An efficient approach to image segmentation is proposed on the basis of 2-D histogram of images, to achieve optimized segmentation effect on grey-level histograms. The algorithm proposed iteratively minimizes a weighted sum-of-squared-error objective function, thus optimizing finally the image segmentation of gray levels on the histogram. With regard to the effect of image segmentation on 2-D histograms, this method is similar to either Otsu's method or entropy's method, but has much more advantages in time occupancy. Therefore, this method has more applications, especially in real-time recognition tracking and surveillance systems.