一种改进的自适应肤色检测算法

Improved Adaptive Algorithm for Skin Color Detection

针对H su R L的肤色检测方法的不足,提出了一种改进的自适应肤色检测算法。用白平衡算法取代了H suR L的光线补偿方法,从而较好地解决了由光照引起的图像彩色偏移,使得检测结果更加稳健。在H su R L的统计模型的基础上,提出了一种自适应检测算法,该算法根据肤色值在非线性变换后的色度空间的二维直方图分布,自动确定肤色模型的位置,从而能更加准确地描述肤色在一幅图像中的分布情况。并将该算法用于彩色视频图像中的人脸检测,实验结果表明,该算法比H su R L的方法更能取得满意的检测效果。

The study is to detect the skin color in color images reliably and effectively. This paper presents an improved adaptive algorithm based on the Hsu R L approach. Firstly, a white balance algorithm is adopted instead of lighting compensation to eliminate the influence of nonstandard illumination, and hence make the result of detection robust. Secondly, based on the characteristic of two dimensional chrominance histogram of the skin color, an adaptive algorithm is proposed. This method can automatically locate the skin color model in chrominance space, so as to describe the skin color patch precisely. Finally, the method is applied in a face detection system for the video sequences. Experimental results show that the approach is more robust and effective than the Hsu R L＇s approach.