摘要
针对已有的图像抠图采样方法易受trimap输入的影响且精确度不足的问题,提出一种基于模糊连接度的抠图样本集构造方法.通过计算模糊连接度求解未知像素到前景边界和背景边界的最强路径,以与最强路径关联的已知像素为中心搜集邻近的已知像素,并构造出未知像素的样本集,且当新的用户笔画加入后,能够快速地更新样本.实验结果表明,文中方法对trimap的依赖性小、采样精确度高、鲁棒性强.
In this paper a color sampling approach based on fuzzy eonnectedness for image matting is presented. Firstly, the strongest paths between unknown pixels and boundary of foreground (or background) are calculated using fuzzy connectedness. Secondly, the known pixels associated to the strongest paths are found. Finally, centered on the associated known pixels, adjacent known pixels are collected to construct sample sets of the unknown pixels. Furthermore, the method can rapidly update samples when new strokes drawn by user are added. Comparative experiments on a variety of input images with different trimaps are presented which demonstrate the accuracy and robustness of the proposed method.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2010年第7期1194-1200,共7页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(60673027)
国家自然科学基金国际合作项目(60811140344)
中央高校基本科研业务费专项资金
武汉大学博士研究生自主科研基金
关键词
图像抠图
颜色采样
模糊连接度
最强路径
image matting
color sampling
fuzzy connectedness
the strongest path