融合双肤色模型及AdaBoost算法的人脸检测

Face Detection Fusion of Dual Skin Models and AdaBoost Algorithm

为提高人脸检测的精度,提出一种融合双肤色模型与Adaboost算法的人脸检测方法。首先采用YCbCr颜色空间的固定阈值模型初次分割图像,利用分割结果修正高斯肤色模型的参数并对图像进行肤色二次分割,对两次分割的结果进行逻辑运算,粗定位人脸区域。结合Adaboost算法,实现对候选人脸区域的精确定位。实验结果表明,该方法提高了人脸检测率,降低了误检率,具有较好的鲁棒性。

In order to improve the accuracy of face detecton, a novel face detection method fusion of dual skin models and AdaBoost algorithm was proposed. Using a fixed threshold skin model in the YCbCr color space, the image segemention of skin region was firstly got. The results correctted the parameters of Gaussian skin color model and image segmentation of the skin region was secondly performed with it. The logical operation was computing with the twice results of skin segmentation which bringed about the coarse positioning face region.Combining Adaboost algorithm, the accurate candidate face region was acquired. The experimental results show that this method improves the face detection rate , reduce the false detection rateand has better robustness.