AN OBJECT SEGMENTATION APPROACH BASED ON UNITED COLOR MODELS FOR SOCCER VIDEO

摘要：

In this paper, an object segmentation approach based on united color models for soccer video is presented. Unlike most existing segmentation measures that only rely on single color model representation, we choose two color models to implement our object segmentation algorithm. The HSI primary model and RGB assisted color model are selected to be complementary to each other, and fusing of the result effectively prevent drifts from the true dominant color statistics, thus we can obtain high pixel-wise segmentation accuracy with these color models. The experimental results on six soccer video clips confirm the effectiveness and robustness of our framework. From the comparative analysis, we obtain a better result than the single HSI color model segmentation measure; especially insure the object connectivity and integrality.

关键词： Color model Dominant color Object segmentation

作者: YAO-QUAN YANG NA LIU WEI GU XIN-YAN LIANG

作者单位: Faculty of control science and engineering, North China Electric Power University, Baoding 071003, China

会议类型: 国际会议

会议名称: 2006 International Conference on Machine Learning and Cybernetics(IEEE第五届机器学习与控制论坛)

会议地点: 大连

会议语种:英文

页码: 3960-3963

在线出版日期: 2006-08-13（万方平台首次上网日期，不代表论文的发表时间）