Research And Implementation Of Realistic 3D Face Modeling

The human face is playing the decisive role in the emotion communication. With the high rapid development of digital and hardware technology, there exists an urgent need to more quickly create more realistic three-dimensional face model. The paper integrates with physiological basis of facial structure, and defines facial feature points set; research then implement feature extraction of human face image, texture blending and mapping; propose the face-specific generation method which bases on standardized data. The main contents of this article:1. Implement Definition of feature points. Reference to the physiological characteristics of human facial structure and MPEG-4 standard FDP face, combined with the practical needs of reconstruction, we set to take 54 points as the set of facial feature points.2. Feature extraction. Improve and simplify two-dimensional picture of the feature extraction implementation process, using cluster analysis and image binaryzation, combined with integral projection method, achieve the facial feature regions and feature points of automatic calibration.3.Propose Overall deformation. After preprocessing CT slice data, three-dimensional facial is reconstructed with MC algorithm, combined with shape of reconstructed model and two-dimensional picture of the measurement data to adjust the generic face model dimensions and positions of facial features of the method, to achieve overall deformation of universal face model.4. study Local deformation of human faces. Universal Face Model and Reconstruction Method are aligned by ICP Registration algorithm, registration results based algorithm using DFFD generates local deformation of face model to adjust, facial model can be viewed as the general characteristics of the generated face model by adjusting registration repeatly.5. Texture mapping of 3D human faces. Improve gaussian-laplacian image decomposition to remove three people face are two-dimensional picture of the fight side seams, and the integration of generated texture map to take the whole view while using cylindrical mapping method to complete the realistic three-dimensional face treatment.6. System implementation. Design and implement a realistic model system for human face, The system integrates several functions which include feature points of two-dimensional photo manual calibration and some feature points of calibration, image stitching and fusion,CT slice data reconstruction, general face model registration and deformation to produce realistic three-dimensional face model and so on. This research is supported by National Natural Science Foundation of China(60736008).