And ro id手机中人脸检测算法的改进及实现

摘要

In order to improve the shooting of smart phones,a new face detection algorithm is presented based on skin color and template.Firstly skin regions are segmented by skin-color clustering in a new color space-YCgCr.Then image binaryzation is completed referring to the built three-dimensional lookup table.The candi-date regions are determined by noise processing and geometric methods,the scale and gray of which are normal-ized afterward.Finally the weighted Euclidean distance is calculated between the candidate face region and the normal reference template,according to the candidate face determined or not.The experimental results show that face detection based on this algorithm has better practicality and efficiency in comparision with Android 2.3 system,and improves the shooting of Android smart phone.%为提高智能手机的拍摄效果，提出了一种基于肤色和模板的人脸检测算法：首先，在一种新的颜色空间---YCgCr中利用肤色的聚类分割出肤色区域；然后，构建三维查找表并据此完成二值化，通过噪声处理和几何方法确定出候选人脸区域并对其进行尺度和灰度归一化；最后，通过计算候选人脸和标准人脸模板之间的加权欧氏距离来度量二者的匹配程度，并依据阈值判断出最终的人脸。实验结果表明，本算法的人脸检测比Android 2.3系统自带的人脸检测具有更好的实用性和高效性，提高了Android手机拍摄的效果。