Region-based hierarchical tongue feature extraction

Tongue diagnosis is important in the Traditional Chinese Medicine (TCM). When doctors analyze the diseases, they usually observe the patient health base on the shape, color and size of tongue. In order to decrease the mistake of subjective judgment, we aim to develop an automatic tongue feature detection system that can help the doctor diagnosis in more scientific way. In this work, a novel region-based hierarchical filtering framework is proposed to robustly detect tongue features. Experiment results obtained using 14-subject dataset with the detection rate being above 70% shows the effectiveness of our proposed framework.