一种快速实现人脸定位与几何校正的方法

A Fast Approach to Face Location and Normalization

提出了利用边缘点集的协方差矩阵的特征值与特征矢量作为人脸图像尺度与方向的粗估计方法,从理论和实验上证明了该方法的可行性.以此方法作为人脸图像精确定位的前期处理,可以大大降低定位的复杂度,从而快速、准确实现人脸图像的定位与几何归一化.实验结果验证了该方法的有效性.

A new method for human face image scale and direction estimation is proposed using covariance matrix eigenvalue and eigenvector of edge set, which is proved by theory and experiments. Based on the method, a novel approach to face location and normalization is presented, which is proved to be fast and precise.