In the paper, a method of image segmentation by ant colony algorithm based on two-dimensional gray histo-gram is presented. The method introduces a ant colony algorithm into image segmentation in order to cluster the gray values, neighboring average gray values, frequencies of gray values in the two-dimensional gray histogram of a image. Aiming at the shortcoming ofmass calculating of a ant colony algorithm, precise cluster centroid locations are gained by the best projection of a two-dimensional gray histogram. A cluster radius, pheromone and a heuristic function are modified for app lication. The experimental results show the algorithm is an effective method.