基于肤色信息和模板匹配的人脸检测与提取

Face Detection and Extraction Based on Skin-color Information and Template Matching

随着人机交互的日益发展,人脸识别中的重要环节人脸检测得到了越来越多的重视。人脸检测的关键在于准确率和效率。针对肤色信息进行人脸检测速度快、易于实现,但准确率不高的特点,将肤色信息与模板匹配的方法结合起来进行人脸检测。利用在YCb Cr色彩空间中的肤色高斯模型,进行肤色区域的分割,经过膨胀腐蚀、长宽比与欧拉数等规则的处理,完成人脸区域的粗检测。然后通过模板匹配的方法进行人脸的细检测,将检测出的人脸区域圈起来并进行提取,形成一系列的人脸图像。实验结果表明,两种方法结合后检测准确率有所提高,速度较快。

With the development of human-computer interaction, face detection, an important part of face recognition, has got more and more attention. The key of face detection is accuracy and efficiency. The speed of face detection by skin-color information is fast and it is easy to be implemented, but the accuracy is not high. A method combined skin-color information and template matching is put forward for face detection. First of all ,the skin-color Gaussian model in YCbCr color space to segment the area of skin is used. After the processing of expansion corrosion,aspect rating,Euler number and other rules,the rough face detection is completed. Then fine face detection by template matching is accomplished. Face region is circled and extracted to form series of face image. Experiments show that the accuracy of the combined method is higher and the speed is faster.