摘要
人体肤色分布在RGB空间非常小的范围内,利用该特征可以将肤色信息从复杂的背景中提取出来。为了有效地去除背景中与肤色相似的干扰成分,我们利用手运动为主运动的特征,通过运动聚类,将手区域提取出来,再对手区域进行肤色滤波可以有效地分割出手势。实验结果表明该方法运算量小,能够有效地进行图像分割,可以用于实时图像识别系统中。
In this paper, a multiple cues (color and motion information) fusion technology is applied to segmentation procedure. In the motion analysis, frame difference and optical flow are used, and then clustering is used to localize gestures. In the segmented image areas, we extract skin-color object respectively in normalized rgb-space and YCrCb-space. The results show that YCrCb-space is suitable for skin-color filtering, and the arithmetic is simple and robust to illumination, it can segment image effectively.
出处
《山东英才学院学报》
2006年第1期56-60,共5页
Journal of Shandong Yingcai University
关键词
肤色
运动信息
图像分割
skin-color
motion information
image segmentation
