基于单视图的多姿态人脸识别算法

针对基于多视图的多姿态人脸识别方法的缺陷，即需要对每个人脸拍摄多个视图为前提条件，提出了基于单视图的多姿态人脸识别技术，首先基于二元高次多项式函数最小二乘拟合方法由单视图通过变形生成多姿态人脸图像，然后基于该单视图和生成的多姿态图像进行多姿态人脸识别。实验结果表明该文算法识别的正确率远高于经典算法。

Multi-Pose Face Recognition Based on a Single View

A multi pose face recognition algorithm based on a single view is proposed in the paper. It consists of two steps. At first step, multi pose face images are synthesized by the image warping from a single view based on a least square fit with a polynomial function. Face is first represented by a dominant point set. Then, the variance of the dominant point set between different poses is fit with a polynomial function and a global morphing field is formed. Finally, multi pose face images are synthesized by image warping from a single view based on the global morphing field. The experiment results show that the synthesized multi pose face images are very similar to corresponding real ones. At second step, multi pose face recognition is performed based on the training set that consists of the single view and the synthesized multi pose images. With the pose changing gradually, the relativity between the corresponding face images at different pose reduces rapidly. So, a hierarchical face model with the division of the face poses and fusion decision are adopted in this section. It first divides face pose space into several pose subspaces and all training samples into several classes according to their corresponding pose. Every hierarchical face model consists of several typical pose. The fusion decision face recognition consists of candidate pose determination of the input face, face recognition in every candidate pose and fusion decision based on the results gotten in every candidate pose. Because face recognition in every candidate pose only search corresponding training images and reduces search space, its computation is cut. The experiment results show that the performance of the algorithm discussed in the paper is by far superior to that of the traditional method.