3D face recognition based on feature detection using active shape models
Sangjun Park, Dong-Won Shin · 2008
This paper represents a method that detects facial features and normalizes 3D range images and 2D images for full automatic face recognition. The active shape models or ASM is applied to extract the position of the eyes, the nose and the mouth. The approximate position of the face in the image is detected using the projection method and the facial profile on the 3D range image so that the initial position of the ASM model is set on the image before the ASM searching. The 3D range image is rotated facing front view using feature points. Then, cropped inside of the sphere to remove none facial parts. The face data is rearranged to fit on the desired frame which size is 201times151. The shape model is built up of 50 images from 10 individuals and the face recognition system is evaluated on 300 images from 30 individuals. The PCA-based hybrid classifier is applied to design the face recognition system with good results.