基于Hausdorff距离的相似性和对称性度量及在人脸定位中的应用

本文提出了一种基于改进的Hausdorff度量的人脸定位方法.首先,改进了模板匹配中基于Hausdorff距离的相似性度量.传统的Hausdorff距离仅仅计算了两个点集之间的平均失配程度,而改进的方法同时考虑了点集之间的距离平均值和标准差,所以对人脸定位中姿态、表情、个体的变化具有更好的鲁棒性.另外还将Hausdorff距离应用于测量对称性这一人脸的重要特征,并将对称性度量和相似性度量结合起来进行人脸定位,从而取得了较高的定位正确率和定位精度.

Hausdorff Distance Based Similarity Measure and Symmetry Measure and Their Applications in Face Localization

An improved approach for face localization based on the Hausdorff Distance(HD)measure is propesed in this paper. First,the similarity measure in the template-based matching is modified.The common form of HD only computes the average of the mis- match of each point in the point set A to the point set B.it has been improved by considering the average and the standard deviation to- gether.The new measure is more robust against the variance of poses,expressions and identities.Then the HD measure is applied to com- pute the symmetry of the face image,which is an important characteristic of human faces.This symmetry measure is combined with the similarity measure during face localization process.The experimental results demonstrate that improvements on both correctness and ac- curacy of face localization.