There are 3 quintessential adaptive thresholding methods,such as maximum variance,maximum entropy and minimum error,which have a large difference in segmentation results of ferrographic images.In order to explain the reason of the difference and discuss their indications to ferrographic image segmentation,the algorithm principles of the 3 adaptive thresholding methods were theoretically analyzed,and the segmentation results of 30 ferrographic images by the 3 adaptive thresholding methods were compared.The conclusions are as follows.Different objective functions are adopted by the 3 adaptive thresholding methods,which bring out different thresholds.For the ferrographic images with bimodal distribution,the three methods have similar segmentation results and the accuracy of the maximum variance method is slightly higher.For the ferrographic images with approximate unimodal distribution,the minimum error method has the best segmentation result,and the maximum variance rule cant segment the abrasive correctly.Therefore,the corresponding adaptive thresholding methods should be selected according to the gray distribution characteristics of ferrography image when analyzing the ferrographic images.