摘要
提出一种使用单幅人脸照片进行特征提取、标准模型变形的全自动三维人脸重建方法。使用改进ASM方法自动精确提取人脸特征点,通过使稀疏形变模型匹配平面特征点来获取照片人脸的深度信息,再将一般人脸模型变形到特定人脸。基于肤色模型优化的ASM提取人脸特征,使得一定角度的侧面照片也可以有很好的重建效果。同时,使用基于肤色模型的纹理融合技术使侧面信息缺失的问题得到很好解决。实验证明,该方法快速简便,只用单幅照片全自动化完成重建,无须用户交互,生成的三维模型有较好的真实感。
This paper proposed an automatic 3D face reconstruction approach based on a single image for feature extraction and standard model deformation. Firstly, the improved ASM method could automatically get the 2D face feature points, and the fitting sparse morphable model could get depth information of the given image. Then deformed a generic model to particular face model. By using ASM facial feature extraction based on color model optimization, the side face image could also get better results, and the texture fusion technology based on color model was very good to resolve the problem of side face information missing. The approach had the merit of using only one image and one generic model, which not only avoided ：onerous preprocessing procedure, but also reduced the computing cost of memory and time. The quality of side face texture was also improved, which strengthened the reality of the synthesized model. Experiments show that this approach is fast and efficient, the result is realistic, and can be well used in practice.
出处
《计算机应用研究》
CSCD
北大核心
2009年第10期3998-4000,共3页
Application Research of Computers
关键词
三维重建
形变模型
一般模型
纹理映射
肤色模型
3D reconstruction
morphable model
generic model
textare mapping
skin model
