In order to overcome the limitation of image segmentation methods based on 2-D histogram and modality in threshold techniques, a method based on the combination of the histogram concavity and the sum of the neighborhood differences was proposed. In addition, the comparative experiment was done when the proposed method and 2-D histogram methods were applied in segmentation on X-ray images of chestnuts, apples and kiwifruits. The results showed that the image segmentation error of the proposed method was less than 2.1%, and its biggest segmentation error was only 23.7% of that of 2-D histogram method. The proposed method could get the fruits’images more precisely.