The concept of the two-dimensional bound histogram (TDBH) is proposed. It is the two-dimensional histogram bound by some prior knowledge, and it can make some image processing methods simple and feasible. Furthermore, an image segmentation method based on the TDBH is presented. In the method, the bound set of an image and its corresponding TDBH are constructed, and the gray-level threshold for segmentation is determined according to the maximum entropy of the TDBH. To show the validity of the method, it is used in segmenting of the image of a small underwater target. The image segmentation method based on the maximum entropy of the one-dimensional bound histogram (ODBH) is also used. The results show that the former, the proposed method, has better antinoise performances and segmental effects than the latter. The concept of the TDBH is significant in theory and engineering. The proposed method is applicable to the image where there is some prior knowledge.