Study on the Algorithm of Cranio-facial Reconstruction
|School||Zhejiang University of Technology|
|Course||Applied Computer Technology|
|Keywords||facial reconstruction feature point extraction laplacian deformation similarity measure|
Computer-Aided Cranio-Facial Reconstruction(CFR) aims at reconstructing the 3D face model from the skull data. It is widely used in many applications, such as the reconstruction of the corpus in archaeology, identification of victims from the corrupted corpus, and face-lifting surgery, etc. In this paper, we present some approaches in CFR, and methods include automatically feature points definition, mesh deformation based on laplacial-coordinate, and computation of the similarity of different 3D models. We summarize our approaches as follows.(1) The automatic selection of feature points. We define the corresponding feature points on the face model and the skull model. The model is aligned into normalization one by using Iterative Closest Point(ICP). We use the shape and curvature index to describe the vertices, and we can find the corresponding feature points on the reference model.(2) Cranio-Facial Reconstruction based on Laplacian deformation. We first obtain the template model of the skull, and we then extract the corresponding feature points on the facial model and skull model. After comparing the difference between two models we final use mesh deformation based on laplacian coordinate to deform the reference facial model.(3) The measure of similarity of cranio-facial models. Firstly, the ICP registration approach is applied into the two models, so we can get the whole similarity of the two models. Then we mark the corresponding feature points on them, and select the area of eyes and noses as local feature, and we get the results according to global and local similariy.(4) A prototype system of computer-aided facial reconstruction. Based on the approaches above mentioned, we implement a prototype system of computer-aided facial reconstruction- FaceMaster, our FaceMaster validates the above approaches .