The maximum entropy segmentation algorithm is good at segmenting images with fuzzy boundary between the target and background, but cannot deal with the image's edge effectively. The maximum interclass variance segmentation algorithm can well identify image edges, however, it cannot accurately segment the image with fuzzy boundary between the target and background. In order to deal with these problems, we propose a maximum entropy image segmentation algorithm based on maximum interclass variance. The algorithm can both segment the image with fuzzy boundary between the target and background and identify the edge of the image effectively. Experimental results prove that the proposed algorithm is superior to the traditional maximum interclass variance algorithm and the maximum entropy algorithm for the image segmentation with fuzzy boundary between the target and background with strong edge recognition ability and better effectiveness and robustness.