Abstract:

BACKGROUND: Traditional Chinese medicine glossoscopy observes tongue image by vision, and to diagnose diseases by physicians’ clinical experiences, which lacks of objective diagnostic methods and criteria, finally resulting in affecting the effectiveness of this diagnostic method. Therefore, it is necessary to realize quantitive and objective glossoscopy using techniques such as computer vision. Tongue body segmentation is the premise of the tongue image recognition diagnostic system. The quality of segmentation directly affects the sequent work. At present, there are many methods of tongue body segmentation, but obtained results cannot ensure the accuracy, and the anti-interference ability is poor.
OBJECTIVE: To design a new method of tongue image segmentation, and effectively isolate tongue body.
METHODS: According to tongue-images gray and color’s features, a segmentation method of combining gray projection and threshold-adaptive way is presented in this paper. By this method, the intensity image of a colorful tongue-image is firstly projected at horizontal and vertical direction to locate the tongue-body area, and then threshold-adaptive way Otsu is used to segment the located area.
RESULTS AND CONCLUSION: Tongue-body segmentation is the basis of automatic tongue-images analysis. The following 281 tongue-images segmentation experiments show the efficiency of this method