Research And Implementation Of 3D Face Reconstruction Algorithm Based On Multiple Photos

In recent years,with the rapid development of 3D reconstruction technology,the 3D face reconstruction technology has been widely used in the fields of face recognition,medical diagnosis,video animation and so on.To restore the human face real,facial and facial key information extraction plays a decisive role.Due to the complexity of human face geometry and facial texture structure,face facial data access to information at a higher cost.Most of the current method is to scan the laser scanner to get facial data information.However,the way to obtain data using scanner is more demanding for subjective factors such as environmental factors and pose of the modeling object,the steps are complicated and the implementation is more difficult.If it is through the detection of two-dimensional images of the face,and the key information extracted on the image way to reconstruct the 3D face.This approach will be implemented in a very convenient way,the cost will be greatly reduced.Users only take a few photos of their own,the photos uploaded to the reconstruction system will be able to truly restore their own 3D face model.At present,many researchers are studying the key point information extraction from 2D images for 3D reconstruction.This session will also start from the work in this area,by researching the key point information extracted from the multi-faceted two-dimensional photos of a certain user,and performing the 3D reconstruction on the basis of these key points.In this session,we mainly study the current mainstream face feature extraction methods and 3D reconstruction techniques,and propose a method based on the facial feature point independent deformation model for 3D reconstruction.Specific work includes the following three aspects:(1)in the 2D face photo extraction facial key points.Face alignment and feature point positioning are done through the improved Supervised Descent Method(SDM),and 83 related feature points are extracted.(2)According to the information of the characteristic points,the deformation control is carried out on the standard 3D model.The deformation process is divided into two parts: the global adjustment and the partial adjustment.The global adjustment mainly adjusts the global 83 feature points of the face.For the local adjustment,interpolation is performed by two or more feature points.A interpolation method based on weight gradient B-spline interpolation is proposed to complete the interpolation work and adjust the personalized 3D face model.(3)extract face texture information from three face pictures(one positive face and two side faces at 45 degrees).On the front face and side face texture of the joint part,through the Poisson fusion way to achieve positive side face texture fusion.The texture information is mapped to the three-dimensional face model,to truly restore the 3D face.Experimental results show that this method of autonomous model deformation can achieve better reconstruction results when performing 3D face reconstruction.