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

With the development of 3D imaging technology,3D face models have been widely used in walks of life,especially in scenes that require high facial details such as film and game productions.Directly creating a 3D face model is complicated and time-consuming.Therefore reconstructing a 3D face model from 2D images has become a current research hotspot.This thesis studies a 3D reconstruction algorithm based on 2D images from three viewpoints.Aiming at the low accuracy of facial feature points extraction,this thesis proposes a grouped facial feature points extraction method,taking the extracted 68 feature points as the basis of model deformation.Aiming at the excessive calculation of feature points deformation,a parallel motion feature points deformation algorithm is proposed,and the 2D feature points are mapped to the 3D standard face model to obtain a rough face model.Then the rough model will be refined.The energy function of the rough model is re-expressed using priori constraints,and then the improved point and curve deformation will make the refined model more in line with the spatial distribution and detailed information of the real face.According to the 3D face modeling method from coarse to fine,each stage of face modeling is sub-divided,and high-precision alignment is performed at each stage to make the model have better effects.Through the parallel deformation of the grouped feature points and the refined processing of the rough model,the texture information and spatial area distribution of the model are refined,so that the reconstructed rough model becomes a high-precision model rich in face details information.The experimental results show that the face model obtained by the method in this thesis has high accuracy and similarity,and can be applied to scenes that require high accuracy and details such as film and game productions.