Summary: | 碩士 === 國立中央大學 === 資訊工程研究所 === 84 ===
We have proposed a facial image processing/recognition/animation system. The research topics of this thesis are parts of this system. There are four topics in this thesis:(1) 3-D facial image conjugation, (2) 3-D face-model recasting and rectification, (3) texture mapping, and (4) facial animation. At first, we use a digital camera to capture personal images with four views, stretch these images by cylindrical projecting method, and then conjugate all stretched images to obtain a cylindrical facial image. In face-model recasting and rectification, we define several feature on 3-D face model and facial image. According to displacements of corresponding feature vertices between model and images, we adjust locations of feature vertices and use interplation method to rearrange non-feature vertex locations. If the recasted face model is not proper, we also provide a 3-D model rectification sub-system. In this sub-system, user can use sensing gloves to modify the face model immediately. In texture mapping, we introduce a texture mapping method for 3-D face model. In facial animation, we adopt abstract-muscle method for simulating facial animation.
The proposed system is created in virtual environment; it supports several equipments for virtual environment. Users can see the face model from arbitrary angles and modify the face model at random by sensing gloves.
|