基于YCgCr颜色空间的不良图像肤色检测

摘要

For low recognition rate of image filtering system on Internet at present, we present a method of objectionable image skin colour detection based on YCgCr colour space.Firstly, possible face region in the image will be detected, and the pixels of complexion are used to obtain human skin colour.Secondly, for those images of face excluded, the skin model constructed off - line will be used to implement skin colour detection.Experiment results show that the proposed method can achieve better result of improving skin colour detection rate and background detection rate for objectionable image with different illumination and in complex background.%针对目前互联网图像内容过滤系统识别率低的情况,提出了一种基于YCgCr空间的不良图像肤色检测方法.首先检测图像中可能存在的人脸区域,利用人脸肤色像素来检测获得人体肤色,其次对不合人脸图像,则利用离线构建的肤色模型来实现肤色检测.实验结果证明,在不同光照以及复杂背景下,该算法能够较好地提高不良图像的肤色检测率和背景检测率.