GUO Jun-bo1, ZHANG Dong-ming1, ZHANG Yong-dong1, et al. Human skin region detection in Web video[J]. 2009, 30(4): 6-10.DOI:
面向互联网视频的人体皮肤区域检测
摘要
提出了一种改进的皮肤区域检测算法
首先以改进的皮肤颜色检测模型为基础检测可能的肤色点
进一步对可能的肤色区域进行腐蚀、膨胀处理以去除零散肤色点
获得连通性较好的可能皮肤区域
之后通过灰度共生矩阵度量纹理特性
去除非肤色区域。实验数据表明该算法能够有效提高视频图像的皮肤区域检测的准确率。
Abstract
An improved detection algorithm was presented.Firstly
an improved skin color model
which adopted many preprocessing including optical offset
auto white balance
and auto gamma correction
was built to detect candidate skin pixels.Secondly
the image was filtered with erosion and dilation to exclude unconnected candidate pixels.Lastly
gray co-matrix was used to exclude those candidate human skin regions with complex texture.Experimental results show that this algorithm improves the skin region detection precision.