摘要: 已有获取显著区域的方法存在不能适应实际物体的大小、包含冗余信息及应用范围有限的问题。为此,提出一种多目标场景下的显著物体提取方法。对基于空间的计算模型得到的显著图进行聚类,将多目标场景划分为多个单目标的子场景,在子场景集合中,引入注意转移机制,并使用基于物体的计算模型依次提取显著物体。实验结果表明,该方法能提取图像中的多个显著目标。
关键词:
显著物体,
场景分割,
聚类,
视觉注意,
注意转移
Abstract: Traditional method of significant area obtaining can not adapt the size of real objects, have redundant information, and the application scope is limited. In order to solve the problems above, this paper proposes an extraction method of salient object in multi-object scene. It clusters the saliency map obtained by space-based model to divide the multi-object scene into several sub-scenes, and introduces a transference of attention mechanism on the sub-scene sets and the object-based model in order to extract salient objects. Experimental results show that the proposed method can completely and correctly extract the multi-object of the images.
Key words:
salient object,
scene segmentation,
clustering,
visual attention,
attention shift
中图分类号:
马志峰, 李颖, 郑芳, 高智勇. 多目标场景下的显著物体提取[J]. 计算机工程, 2012, 38(17): 209-213.
MA Zhi-Feng, LI Ying, ZHENG Fang, GAO Zhi-Yong. Extraction of Salient Object in Multi-object Scene[J]. Computer Engineering, 2012, 38(17): 209-213.