关键词: 图像分割, 阈值, 最小交叉熵, 二维直方图

Abstract: Entropy is acquired by calculating image’s histogram，which only uses gray information of pixels.Minimum cross entropy is to search optimal threshold that can make minimum difference of information between original image and segmented image.On the basis of minimum cross entropy，this paper defines two-dimensional within-class minimum cross entropy by introducing spatial information of image.Moreover，this paper proposes an image segmentation method based on 2D within-class minimum cross entropy.Experimental results show that proposed method can obtain better segmentation effect than minimum cross entropy.In order to promote the speed of segmentation，corresponding recursive algorithm is proposed.Analysis shows that the computation time of 2D within-class minimum cross entropy method is reduced from more than 3 hours to several seconds.

Key words: image segmentation, threshold, minimum cross entropy, two-dimensional histogram