基于动态阈值图像分割法的人脸识别技术研究

The study of face recognition technology based on dynamic threshold image segmentation method

提出一种基于动态阈值图像分割的人脸识别方法.在灰度级别下,基于图像分割中的Fisher准则,利用Fisher函数的类间均值最大、总类内方差最小的原则,自动获取待检测图像所对应的最佳分类阈值,并根据所得的动态阈值进行肤色分割,然后再根据阈值解码器,实现肤色似然图的二值化,得到肤色分割后的二值化图像,从而检测到包含有人脸的肤色区域.实验结果表明,该方法改善肤色分割性能,能够在负载复杂背景下实现肤色区域的精确分割,提高人脸检测的速度和精度.

We proposed a face recognition method on the base of dynamic threshold image segmentation. In the gray level, based on the image segmentation Fisher criterion, following the principles of the Fisher function of the maximum between-class and the general category of minimum variance, it automatically got the best classification of the image on the corresponding threshold value. It could also conduct the skin color segmentation according to the dynamic threshold obtained. Then it achieved the binarization of the skin likelihood figure according to the threshold decoder, and it got the image binarization after skin color segmentation so that the skin color area including the face could be detected. The result of this experiment demonstrated that the method could improve the performance of the skin color segmentation, achieve the exact cutting of the skin color area under the complex load, and improve the speed and accuracy of the face detection.