基于多数据融合的快速人脸检测与特征定位算法

针对彩色图像提出了基于肤色和特征验证的人脸检测算法,由肤色分割、排除假区域、特征验证3部分组成。在肤色分割部分,提出自适应阈值法,再通过几何特征排除一些假区域,运用边缘检测、模板匹配和投影法精确定位眼睛和嘴巴,从而验证确认人脸,最后通过投影法定位出鼻子。实验证明该算法检测速度快、准确率高,具有较好的鲁棒性。

Algorithm for Fast Detecting Human Faces and Locating Features Based on Multi-data Fusion

This paper presents a new algorithm for human face detection of color image based on skin col- or and feature verification.The algorithm consists of three parts:skin color segmentation,false area exculsion, and feature verification.First,an adaptive threshold values algorithm is proposed for segmentation,then the ex- clusion of some false areas are conducted based on geometric features,and finally the face is verified based on the eye and the mouth using edge detection,template matching and projection.The performance of the algo- rithm is experimentally evaluated using a variety of human faces.The results show that the algorithm has high accuracy,robustness and low computational cost.