自适应搜索目标的快速彩色复合人脸检测方法

摘要

针对AdaBoost人脸检测方法搜索时间较长,不利于在手机等嵌入式平台上应用的现状,提出了一种结合肤色分割、人脸几何特征和AdaBoost的自适应搜索窗口和搜索步长的快速人脸检测方法.该算法在HSV颜色空间对图像进行分割,结合人脸几何特征对分割后的灰度图像进行面积滤波.最后提取滤波后的图像轮廓,结合经验系数得到自适应搜索窗口和搜索步长.实验结果表明,自适应算法不仅能检测出不同尺寸的人脸,而且检测速度快,能节省51.17%的搜索时间.%A fast composite color face detection algorithm of adaptive search object is proposed to overcome the disadvantage of the traditional Adaboost face detection way, which needs a lot of time to detect the different size face object. Firstly it segments skin in the HSV color space, then it uses geometrical characteristic to filter the gray image, finally it extracts the facial contour and calculate the size and the step of the search windows. Experimental results show that the method is effective for both detection speed and detection performance. The method can detect the different size face, and it can save the detection time about 51.17％.