Research On Face Segmentation In Face Diagnosis

The traditional Chinese medicine face diagnosis (TCMFD) is the method which doctors mainly diagnose disease by observing the face appearance, color, skin, texture, flaw etc. The color of face is one of most important features in TCMFD. This research aims to TCMFD objectivity and quantification which enables computer to assist doctor in the process of Chinese diagnosis. So how to locate accurately different face regions is the first question to be solved.First we introduced some basic method in face location, and requirements in face diagnosis locating. At the same time, a general description of color diagnosis development is given.For the special requirements of face diagnosis segmentation, we proposed a face diagnosis segmentation algorithm based on the ellipse-gather model which is proposed by Anil K.Jain etc. The size of original picture is 1944×2592, and it wasted much time to skin detection by the ellipse-gather model. Most of background of the pictures is simple, so we propose another fast algorithm.On the foundation of face diagnosis segmentation algorithm, we designed the platform of face segmentation. And we accomplished the function of automatic segmentation, semi-automatic segmentation and artificially segmentation.With the platform we respectively adopt automatic and artificial method to get the blocks in the face based on the theory of TCM. We present the distributions of these blocks in the RGB-color space. Then we combine the k-folder cross validation and knn-classifier to validate the segmentation results. At last we proposed the principles of block size selection.