Research And Implementation Of Cranio-Facial Reconstruction Based On Regional Deformable Model

Computer-aided craniofacial reconstruction(CFR) is a simulation technology to reconstruct facial models, and it is based on the morphological relationship between skull and facial, referred to the tissue thickness of specific populations. It has been widely used in archaeology, forensic investigation, medical plastic surgery, facial expression animation, which is a challenging research and difficulty topic in computer graphics. In this paper, on the basis of craniofacial CT data, the author created CFR knowledge base and selected the best matched specimen types as to recovery guiding knowledge; then decomposed the craniofacial model into segments, such as eyes, nose and mouth and so on; at last deformed each segment to reconstruct the CFR. The main contents are listed as following:1) Created computer-aided CFR knowledge base. The knowledge base can manage effectively the acquired specimen data. We preprocessed the obtained raw CT data and acquired the single outer craniofacial data. Creating the knowledge base, these data and basic information of specimen are stored into the base as recovery guiding knowledge, which is supported for specimen selection.2) Proposed a craniofacial feature point localization method that it is based on the Relative Angle-Context Distribution(RAC) algorithm and Support Vector Machine. Located craniofacial feature point is the basis of craniofacial segmentation and CFR. The method can track the corresponding points to a small neighborhood area and then precise point's location by the Support Vector Machine, which is improved the efficiency and accuracy of feature point's localization.3) Proposed a Global Point Signature(GPS) clustering grid decomposing method which is based on central points initialized. The method calculated the GPS vector representation of model points and took use of K-means clustering, and then decomposed craniofacial models into a few of segments. The method can improve the accuracy of segmentation.4) Proposed the Thin Plate Spline(TPS) partition deformation CFR method and Moving Least Square(MLS) deformation CFR method. For the partition of the local grids, the paper analyzed the principle of TPS deformation and MLS deformation methods, and then selected the recovery guiding knowledge from knowledge base, decomposed the models, located the facial feature points to TPS deformation, at last spliced smooth processing, completed the CFR. The outlook is more realistic than overall MLS reconstruction.5) Designed and implemented of the FacialRec craniofacial reconstruction system. Taking the advantage of Visual C++ and Matlab, the paper designed and realized FacialRec craniofacial reconstruction system, which is favorable for supporting the research.This research is supported by National Natural Science Foundation of China(60736008).