基于肤色分割、区域分析和模板分布的人脸检测研究

提出了一种基于肤色分割、区域分析和模板分布的彩色图像人脸检测算法。首先对输入的彩色图像利用混合高斯模型和亮度模型进行分割,然后根据人脸五官的结构特征对得到的区域进一步分析处理,获得所有可能的候选人脸。接着构造了一种基于双眼和人脸模板的概率模型并利用其对候选人脸进行最终检测。实验结果表明,文章提出的算法具有较高的检测正确率和自适应能力;同时具有快速的检测速度。

Face Detection Based on Skin Color Segmentation,Region Analysis and Template Distribution

In this paper,a face detection algorithm for color image based on skin color segmentation,region analysis and template distribution is presented.First,the image with human faces is inputted and dealt with a Mixture-of-Gaus-sian(for multimodal distributions)model and a luminance model to get all skin color regions.The image obtained can be further divided into several sub-regions,which are then analyzed so as to eliminate some non-face regions according to the facial features and some general knowledge.Consequently,all candidate face regions with special position and size are acquired.In the end,a probability model with the distribution of double eyes and human face are constructed to perform the final face detection.Experimental results demonstrate the high probability of correct detection and adaptive abilities of the algorithm,and at the same time ,a decent speed of the face detection is achieved.