节点文献
基于视觉内容与语义相关的图像标注模型
An Image Annotation and Refinement Model Based on Visual Content and Semantic Correlation
【摘要】 针对当前标注系统的不足,设计了一种高效的标注模型,其标注步骤包括标注和标注改善,标注算法采用加权的正反例标志向量法,标注改善采用NGD方法。实验表明,标注效率远优于经典的标注模型,标注质量优于大多数标注模型。
【Abstract】 The efficiency and qulaity of image annotation system determine the ability to manage images in the fields of computer vision and image retrieval.To overcome the drawback of current annotation system,an efficient annotation system is designed,including annotation and refinement stages by weighted positive and negative symbol vector method and NGD method respectively.The experiments demonstrate our proposed system perfomance,whose efficiency outperforms classicial image annotation models and qulity outperforms most current image annotation models.
【关键词】 图像标注;
标注改善;
归一化Google距离;
【Key words】 image annotation; annotation refinement; normalized Google distance;
【Key words】 image annotation; annotation refinement; normalized Google distance;
【基金】 中央高校基本科研业务费专项资金项目(DC10040111);辽宁省教育科学“十二五”规划立项课题“应用型院校中本科生研究性学习模式的研究与实践”(JG11DB062)
- 【文献出处】 大连民族学院学报 ,Journal of Dalian Nationalities University , 编辑部邮箱 ,2012年01期
- 【分类号】TP391.41
- 【下载频次】99