• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Conference on Universal Digital Library(ICUDL2005); 20051031-1102; Hangzhou(CN) >Automatic Image Annotation Based on Normalized Relevance Model with Annotation Order
【24h】

Automatic Image Annotation Based on Normalized Relevance Model with Annotation Order

机译：基于带标注顺序的归一化相关模型的图像自动标注

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Efficient image retrieval is necessary for digital library. There are many successful techniques in image retrieval. However, due to the semantic gap between low-level features and high-level semantics of images, these techniques can't achieve satisfied performance. In order to fill up this semantic gap, statistical model is introduced to annotate semantic content to images automatically. Such model views the process of annotating images as translating a visual content language to a textual language and usually encounters the problem of skewed distribution of words frequency. Since words of low frequency in such distribution are often used as queries, the efficiency of semantic images retrieval is degenerated. In this paper, we propose a new model to annotate images with a normalized words frequency distribution. We take the words-sequences' and low-level features' order-information into consideration and improve the efficiency of annotation. Experiments show that the combination of statistical model and order-information lying in the annotated image set achieves good performance and provides us a new way to effectively annotate the unknown images.
机译：对于数字图书馆，有效的图像检索是必要的。在图像检索中有许多成功的技术。但是，由于图像的低层特征和高层语义之间的语义鸿沟，这些技术无法获得令人满意的性能。为了填补这一语义空白，引入了统计模型来自动将语义内容注释到图像中。这种模型将注释图像的过程视为将视觉内容语言转换为文本语言的过程，并且通常会遇到单词频率分布偏斜的问题。由于这种分布中的低频单词经常被用作查询，因此语义图像检索的效率降低了。在本文中，我们提出了一种新的模型来对具有标准化词频分布的图像进行注释。我们考虑了单词序列和底层特征的顺序信息，提高了注释的效率。实验表明，将统计模型与顺序信息结合在带注释的图像集中可以达到良好的性能，为有效地对未知图像进行注释提供了一种新途径。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号