基于流行排序的多示例图像检索方法.pdfVIP

计算机工程与设计 Computer Engineering and Design ( ) 虎晓红，钱旭，王珂：基于流行排序的多示例图像检索方法 2009,30 21 5007 开发与应用 基于流行排序的多示例图像检索方法 1,2 2 3 虎晓红 ， 钱 旭 ， 王 珂 (1. 河南农业大学信管学院，河南郑州450002 ；2. 中国矿业大学机电与信息工程学院，北京 100083 ； 3. 首都医科大学 生物医学工程学院，北京 100069) 摘 要：为了提高图像检索的性能，提出了一种基于流行排序的多示例图像检索方法，将分割后的图像表示为多示例的形 式，通过给出适合图像在包空间的度量方式，有效结合流行排序和多示例学习的方法来进行图像检索。实验结果表明，采用 所提出的方法的检索结果与传统的检索方法相比，检索率得到了明显的提高，检索结果更符合人的视觉习惯。 关键词：图像检索; 流行排序; 多示例学习; Hausdorff 距离; EMD 距离 中图法分类号：TN911.73 文献标识码：A 文章编号：1000-7024 (2009) 2 1-5007-03 Manifold ranking approach to multi-instance image retrieval 1,2 1 3 HU Xiao-hong , QIAN Xu , WANG Ke (1. School of Information and Management Science, Henan Agricultural University, Zhengzhou 450002, China; 2. School of Mechanical Electronic and Information Engineering, China University of Mining and Technology, Beij ing 100083, China; 3. School of Biomedical Engineering, Capital Medical University, Beij ing 100069, China) Abstract ：To improve the performance of image retrieval result, a multi-instance retrieval approach based on manifold ranking is pro- posed. The segmented image is represented under multi-instance learning framework. A novel graph based learning approach to image retrieval using multiple instances is presented, which extends the conventional manifold ranking to multi-instance setting by introducing the adaptive geometric relationship between two bags of instances. The experiments over Corel