Taipei：Asia-Pacific Education &Research Association
We propose a simpler and faster method to construct 3D face model. First, we extract ASM face features and get the depth face image with Kinect. Then, we describe the depth face image with OpenGL in three-dimension coordinate. Owing to the dropped depth information in face edge, we use CANDIDE-3 face model as basic face skeleton to remedy the distortive profile. Here, we match the CANDIDE-3 face model and front face image with the coordinate of ASM feature points. After we get an incomplete 3D face, we need to fill the data to the empty part of side face. According to the skeleton coordinate of face model, we extend the edge information for each row in profile face with the edge pixel in side face till the end edge of face model. Because the repaired 3D face still retains many texture information of original front face image, we can easily achieve a satisfied result with any rotation angle from 0 to ± 90 degree.
Proceedings of International Research Conference on Information Technology and Computer Sciences, pp.32-39