摘要
Grow Cut算法是基于细胞自动机的交互式图像分割方法,针对该算法要求用户标记初始种子需要较多工作量,且带有一定的主观性和不确定性,导致分割结果出现较大误差的问题,文中提出了简化标记,自动生成初始种子模板的基于标记提取的Grow Cut分割算法。该算法在Grow Cut算法基础上通过阈值和形态学方法预处理生成初始种子模板,运用细胞自动机迭代算法完成目标的提取。算法避免了用户人工交互约束的繁琐操作,实现了完全自动分割。通过实验对彩色图像进行自动分割,实验结果证明该算法简便、用时少,分割结果比较精确。
Recently, the GrowCut algorithm majoring in interactive image segmentation, and a common problem of GrowCut is the initialization requires distributed seeds with subjectivity and uncertainty by users. This paper develops an automatic object segmentation that provides effective and robust segmentation of color images by GrowCut based on marker extraction. The process is iterative, utilizing thresholds and morphological methods to generate the initial seed template. The main contribution focuses on performing automatic image segmentation which avoids the constraints of user interaction. The superiority of the proposed method that obtains good segmentation result rapidly which is examined through a number of experiments with color images. The experimental result shows that the proposed method gives better performance and less time.
出处
《信息技术》
2015年第5期76-80,共5页
Information Technology
基金
教育部博士点基金(20113227110010)
江苏省高校自然基金(10KJB520004)
江苏省软件与集成电路专项基金(2009[100])
江苏省普通高校研究生科研创新计划(CXZZ11_0575)