基于肤色模型和FloatBoost的多姿态人脸检测

Multi-view face detection using skin color model and FloatBoost

多姿态人脸检测是人脸识别系统必须解决的关键问题之一。利用光照鲁棒的肤色模型来搜索待检图像的可能人脸区域并进行肤色分割,结合分割区域的几何信息确定最终的候选人脸区域,然后对人脸的关键特征进行定位,按规则计算重要特征块的中心,将这些中心点确定的符合条件的候选区域利用FloatBoost进行分类,最终实现了快速准确的多姿态人脸检测。

Multi-view face detection is a key problem for face recognition system. This method searches the potential face regions using the proposed skin color model, chroma map is adopted to obtain the four-connected components from the skin color seg- mentation blocks, it labels them, and identifies the center of each block, the faces verification is performed through the classifier based on FloatBoost.