A Face Pose Estimation Method for Virtual Try-On System

In the virtual try-on system,virtual objects need to be superimposed on the real-time face image accurately,which determines the experience of the virtual try-on system.The key technology is to estimate the real-time 3D face pose quickly and accurately.In response to this technical requirement,a method for real-time estimation of the face pose was proposed,and a virtual try-on prototype system was constructed.The method first used a depth camera to collect face data,analyzed the color and depth information of a single frontal image,and calculated the 3D face feature points;then combined the 2D real-time face feature points extracted from video frames,and used the direct linear transformation method to solve the real-time face pose.Based on the consumer-level depth camera Astra Mini and Unity3D engine,a prototype system was developed and a virtual try-on experiment was conducted;the accuracy of the pose estimation method was tested using the Biwi data set,and the robustness was tested under different lighting and partial face occlusion conditions.The results show that:the real-time performance,accuracy and robustness of the pose estimation method in this paper can satisfy the experience of the virtual try-on system.