AMGE: A Tongue Body Segmentation Assessment Method via Gradient Energy

Tongue body segmentation is essential for the computerized tongue diagnosis. Several segmentation methods have been developed and some evaluation methods have been used for test the segmentation. However, it is difficult to assess the non-contour parts on the results of segmentation by using the existing assessment methods. To deal with this problem, in this paper, we proposed a novel assessment method for tongue body segmentation based on the characteristics of tongue body contour, called AMGE. In AMGE, There are three steps. Firstly, since of the closed circle structure, the tongue contour is converted into polar coordinate. Secondly, based on the characteristics of tongue body contour, we propose tongue body contour energy which contains radial-based energy function and angle-based energy function. Based on this contour energy function, we can evaluate the tongue body segmentations. Finally, one single threshold is selected to detect the non-contour parts on contour which have high values of energy. Experiments show that the proposed assessment method is superior to the conventional area-based methods and boundary-based methods.