Abstract: To avoid the shortcomings that it is imprecise to use feature line to locate the intersection of the adjacent photographic images in traditional texture image generation method，a new algorithm called “geometric model segmentation” is introduced.Through cutting the geometry model of human’s head，the contours of the divided model photos are obtained to manicure the corresponding face photos as the boundary line.The manicured face photos can be used to generate texture picture without overlapping and gap between the adjacent pictures.The cylindrical projection is introduced to construct 3D specific face model through the texture mapping by using of the above texture picture.The experiments show that the new algorithm is an effective way to generate texture images in photo-realistic 3D human face modeling.

Key words: feature line, geometric model segmentation, texture images, cylindrical projection

摘要： 为了克服传统的照片图像拼接方法中利用特征线进行不同照片之间公共交界线定位不准确的缺点，提出一种“几何模型切分”的人脸纹理图像生成算法。通过对人脸几何模型进行切分，以切分后模型图片的轮廓作为边界线裁剪相应的人脸照片，实现不同照片之间交界线的准确对接，并采用柱面纹理映射方法将生成的纹理图像映射到特定人脸几何模型上。实验结果表明，采用提出的“几何模型切分”算法生成的人脸纹理图像进行纹理映射可以得到较好的真实感三维人脸模型，是一种生成人脸纹理图像的有效方法。

关键词: 特征线, 模型切分, 纹理图像, 柱面映射