融合多特征的人脸检测方法

Face Detection Based on Multi-Features Fusion

摘要:

针对复杂环境下基于肤色模型的人脸检测误检率较高以及Adaboost算法对高分辨率图像时间效率低，提出了一种新的结合肤色模型和皮肤纹理特征以及Adaboost级联分类器的人脸检测方法，并改进了基于纹理刷色阶偏差法的皮肤纹理特征提取方法。该算法充分融合了肤色模型简单快捷、皮肤纹理突出的特性以及Adaboost级联分类器检测率高等优点。实验表明，该方法检测率高且有较好的鲁棒性。

Abstract:

In view of the high false detection rate of skin color model and the low time efficiency of Adaboost algorithm for complex environments, a new face detection method is proposed based on the fusion of skin color model, skin texture and adaboost cascade classifier. For the proposed method, the skin texture extraction approach based on the texture-brush color-order deviation is improved. This algorithm successfully integrates the simplicity and quickness of skin color model, prominent characteristics of skin texture with the high detection rate of adaboost cascade classifier. Experimental results prove the high detection rate and good robust of the method.