Tongue image change detection

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] According to the theory of traditional Chinese medicine (TCM), the tongue has connections to different parts of the human body. The status of the tongue may reflect a person's overall health state, as well as the condition of specific organs. This thesis introduces an automatic TCM diagnosis system called ITongue. With this system installed on Apple iPhones, users are able to upload photos of their tongues. The server makes an automatic diagnosis and gives dietary advice accordingly. We propose an improved GrowCut segmentation algorithm. Compared to the traditional GrowCut algorithm, our method is able to label the seeds automatically without user interaction. To make the two input images align, we correct the orientation of tongue images and adopted the Iterative Closest Point (ICP) algorithm for alignment. Each input image is represented by a cloud of points. By translating and rotating the second cloud, we minimize the distance between the two clouds to achieve alignment. Finally, a pixel-wise method for difference detection is performed by computing the Euler distance between each pair of matched pixels. We also apply a patch-wise method and t-test for better results.