三维人脸纹理改善及其在模型匹配中的应用

Improvement on 3D Face Texture and Application in Process of Model Matching

为改善三维人脸库的颜色纹理,基于纹理映射技术,提出了一种改善方法．使用统计学中曲线拟合与线性回归的方法,根据扫描仪获取的三维人脸库颜色纹理空间,拟合、逼近真实纹理空间,进而得到改善的颜色纹理空间,建立了三维人脸库颜色纹理空间到改善的纹理空间之间的映射关系．根据该映射关系改善了三维人脸库颜色纹理,缩小了其与真实纹理之间的差异．实验结果证明,用该方法改善后的纹理与真实纹理的差明显小于扫描得到的纹理与真实纹理之间的差;在模型匹配过程中使用改善后的纹理,可提高模型匹配速度,缩短匹配时间．

A novel method based on texture mapping was presented to improve texture of 3D face database. Curve fitting and linear regression were used to fit and approximate to the realistic texture space based on texture space of 3D face database. According to the improved texture space, the texture mapping rule was established. Then the texture mapping rule was used to improve the texture of 3D face database and to reduce difference between texture and realistic texture of 3D face database. Experimental results prove that difference between the improved texture space and realistic texture space is smaller than that between texture space and realistic texture space of 3D face database. When the improved texture is used in process of model matching, the speed of model matching gets faster with shorter time-consuming.