Automatic tongue location and segmentation

Automatic tongue location and segmentation is challenging due to the complexity of pathological tongues, variance of tongue shapes and interference of lips. This paper proposes an unsupervised tongue location and segmentation method to address these problems. First, look-up-table (LUT) based real AdaBoost algorithm is introduced to get the cascade tongue detector which can make locating tongue in human face images faster. Then, color cancellation is used to remove the lip and a method which combines radial edge detection with active contour model is presented to get the exact contour of tongue. The whole algorithm can be done without any human interference and experimental results are pretty satisfying with a precise rate up to 94.91%.