Research On 3D Face Modeling Algorithm Of Three-View Images

3D face modeling is a research hotspot and difficulty in academic circles.Many researches have realized face modeling of single-view image,but they can’t solve the problem of incomplete information extraction caused by face self occlusion.Therefore,this thesis studies how to use three-view images to build a realistic high-precision 3D face model.(1)Face geometric modeling method of three-view images based on UV position map is proposed.Aiming at the problem of serious vertical stripes in the model of PRN method,this thesis improves and extends it to three-view images,proposes a strategy of 3D face information fusion based on UV position map,designs the weight function for fusion,and realizes the 3D face geometric modeling of three-view images.The experimental results show that this method effectively improves the phenomenon of vertical stripes and increases the modeling accuracy by about 5%.(2)A complete 3D face texture generation method based on UV texture map is proposed.The complete face texture can’t be obtained from single-view image,this thesis uses the Zbuffer to extract the visible texture of three-view images and store it in the UV texture map,then fuses it to generate the complete texture.The barycentric coordinate mapping is used to map the texture to the model to build a realistic 3D face model.Compared with other models,this model has complete texture and higher similarity with the input face.To sum up,the realistic and high-precision 3D face modeling algorithm proposed in this thesis can be applied to scenes such as face animation design,virtual makeup,identity and expression recognition.In this thesis,the applications of face depth map estimation,pose estimation,"face eyebrow replacement" and "face replacement" are realized,which further proves that this algorithm has great practical value.