基于姿态估计的多视角人脸检测

目前，对于多视角的人脸检测仍然是一个具有挑战性的难题。为了解决这个问题，提出了一种基于姿态估计的人脸检测的方法。在训练人脸分类器时，我们将视角范围划分成几个小区间，为每一个小区间利用Adaboost算法训练相应的分类器；在对图像进行检测时，对于任何给定的一幅图像，我们首先对其进行姿态的估计，并根据估计的结果选择相应的检测器对其进行检测。实验结果表明，提出的方法能很好地检测多视角人脸图像，在检测率和检测速度上都有很好的效果。

Currently, the multi-view face detection is still a challenging object. To solve this problem, this paper proposed a face detection method that based on pose estimation. Before training face classifiers, we divided the angle of view into several small areas artificially, and trained the corresponding classifier by Adaboost algorithm for each area; when detected the images, for any given image, we estimated its posture firstly, and according to the estimated result selected the appropriate detector for detecting. Experimental results showed that the proposed method can detect multi-view face images, and the detection rate and detection speed are both very good.