一种改进的人脸纹理映射方法

摘要

Texture mapping in 3D face modeling tends to distortion. In this paper, an improved method for texture mapping was proposed. Firstly, it mosaicked the images by means of column feature matching based on the informa tion of face feature points, and fused the texture image using weighted smoothing algorithm and the classic Laplacian pyramid method. Secondly, a special positioning method was used to calculate the texture coordinates corresponding to 3D face model vertex. Finally, the full view of face texture image was mapped to the 3D face mesh model to get a realistic face model. This method has been proved faster and robust by the experiment, and it can effectively prevent texture image and texture mapping from distortion.%识别人脸图像优化问题,由于人脸纹理映射时容易出现的纹理扭曲和畸变现象,影响图像的真实性,提出一种改进的纹理图像映射方法.根据人脸特征点信息,利用特征块匹配法配准正侧面人脸图像,并使用加权平滑算法和金字塔方法对图像进行双重融合实现图像的平滑过渡；然后利用一种特殊的定位方法来计算三维人脸模型上顶点对应的纹理坐标;最后根据纹理坐标将纹理图像映射到三维特定人脸模型表面上,可得到具有真实感的三维人脸模型.实验结果表明,纹理合成方法简捷有效,纹理映射结果生动逼真,有效的避免了纹理图像的扭曲和畸变现象,具有较高的鲁棒性.