3D Reconstruction Of Multi-view Face Images Based On Morphable Model

In recent years,three dimensional face model is applied to all kinds of fields widely.It is one of the most challenging topic in the field of computer vision and image graphics to establish the realistic 3D face model.Compared to the two dimensional face image,three dimensional face model can reflect more information.It is because three dimensional face model has a wide application prospect and value that 3d face reconstruction has become the most active research topic in the world.At present,many people use the morphable model to reconstruct 3D face,most of which are reconstructed with single image.Because the single image lacks the information of multiple angles,it is not possible to determine that each Angle can get a better effect.To solve this problem,this paper proposes a 3d reconstruction method based on the morphable model using multiple images from different angles.Includes the following:(1)Morphable model and multi-view face images matching is the most important steps in 3D reconstruction.In order to obtain a more accurate matching model,this paper proposes a new morphable model and multi-view face image matching method.According to the feature points of each image,the 3D Dense Face Alignment(3DDFA)method is used to estimate the model parameters of each image.The morphable model is obtained according to the model parameters of image to obtain a set of 3D faces model after deformation.(2)In order to obtain a matching face model,this paper presents a model function.The function is associated with the parameter information of each image,and by obtaining the average of the face model,the 3D model of the target face is obtained.The experimental results show that the proposed model is more precise from qualitative and quantitative perspective,which shows the effectiveness of the method.(3)Although the matched model is similar to the target face,it still lacks some local details.This paper use the photometric stereo technology and restore the surface normal information by assuming a Lambertian lighting model to improve the details of the model.The experimental results are compared with other reconstruction methods,and it is proved that the reconstruction model has better effect and stronger robustness through qualitative and quantitative analysis.(4)In order to put the theoretical work of this paper into practice,a simplified 3D face reconstruction system is designed and developed.The system realizes the basic process of face reconstruction and further verifys the effectiveness of the proposed scheme.