3D Facial Reconstruction Based On Single Frontal Photo

Face is the most important and most direct carrier for human emotional expression and exchange. It has a wide application prospect that generated with a realistic three-dimensional human face through computer. And it is one of the most challenging issues on computer graphics, computer vision, artificial intelligence and other areas in recent years. With the development of computer graphics, realistic 3D face reconstruction has been used in increasingly wide application areas such as film and television, video games, virtual environment, Paramedic and criminal identification. The research of the realistic three-dimensional human face modeling has increasingly important theoretical significance and application values.A realistic three-dimensional face automatic modeling technique that based on one frontal face photograph will be discussed. The input of this method is just a frontal face photograph with Neutral facial expression and uniform illumination. A personalized face mesh will be generated by face recognition, triangular mesh decomposition to the photo in combination with standard three-dimensional human face model. In the final, a realistic 3D face model will be built with the texture features extracted from the picture and the personalized face mesh. The main contents of this article are including:Recognition algorithm of human face based on Sobel-Otsu is a new method. The algorithm uses the eight directions of Sobel operator mug shots of the preliminary outline of human face, and then Otsu algorithm is used to seek the threshold value which we use to get the binary graph.The CEAIP facial feature extraction algorithm is researched. First of all, the area of each facial feature will be found by using the traditional proportion of human face. Then the integral projection map will be drawn by using the integral projection method in each area. In the end, the coordinate of each facial feature is determined by Analysis of the integral projection.A three-dimensional human face model generation algorithm is discussed increasingly. The algorithm uses the Delaunay meshing algorithm to Decomposition the photo, and then substitutes the mesh into the revised standard face model to get personalized 3D face mesh model. And presents a grid surface smoothing algorithm based on the three-point. At last, a realistic 3D face model will be built with materials through the photo.Using visual studio 2005 and CSGL Development Kit to develop the Realistic 3D Facial Automatically Reconstruction System based on the algorithm discussed in the last.