基于肤色模型的人脸检测  

Face detection based on skin color model

作　　者：赵媛媛 余顺园 ZHAO Yuan-yuan;YU Shun-yuan(School of Electronic and Information Engineering,Ankang University,Ankang 725000,Shaanxi Province,China)

机构地区：[1]安康学院电子与信息工程学院,陕西安康725000

出　　处：《信息技术》2024年第7期53-59,共7页Information Technology

基　　金：国家自然科学基金(61801005)。

摘　　要：人脸检测过程中易受到面部细节、形状、表情、姿态等影响,肤色特征作为人脸中最具有识别性的信息之一,具有相对稳定性。文中基于肤色模型实现人脸检测与标记,在YCbCr色彩空间构建肤色模型,通过光线补偿消除肤色提取过程中光照不均匀、亮度变化等环境因素的影响,基于形态学滤波滤除非脸部区域的噪声,结合人脸区域的几何特征最终实现人脸定位并框选标记。在光照不均、有遮挡、大角度等多种情况下,人脸图像测试结果表明,所实现的算法能够较好地实现人脸检测与定位。Face detection process is easily affected by facial details,shape,expression,posture,etc,and the skin color feature,as one of the most recognizable information in human face,is relatively stable.This paper implements face detection and labeling based on the skin color model.The skin color model is constructed in the YCbCr color space,and the influence of environmental factors such as uneven illumination and brightness changes in the process of skin color extraction is eliminated through light compensation,and the noise in the non-face area is filtered based on morphological filtering.Combined with the geometric features of the face area,the face positioning and box check mark are realized.The test results of face images under various conditions,such as uneven illumination,occlusion,and large angles,show that the implemented algorithm can better achieve face detection and localization.

关 键 词：肤色模型 颜色空间转换 光照补偿 人脸几何特征 人脸检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]