...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Journal of Computers >Image Annotation Refinement Using Dynamic Weighted Voting Based on Mutual Information
【24h】

Image Annotation Refinement Using Dynamic Weighted Voting Based on Mutual Information

机译：基于互信息的动态加权投票的图像标注细化

获取原文
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Automatic image annotation is a promising solution to narrow the semantic gap between low-level content and high-level semantic concept, which has been an active research area in the fields of image retrieval, pattern recognition, and machine learning. However, even the most dedicated annotation algorithms are often unsatisfactory. Image annotation refinement has attracted much more attention recently. In this paper, a novel refinement algorithm using dynamic voting based on mutual information is proposed. Unlike the traditional refinement algorithm, the proposed algorithm adopts dynamic weighted voting to measure the dependence between the candidate annotations, which not only permits that the annotations with higher probabilities deny the annotations with lower probabilities, but also permits that the annotations with lower probabilities deny the annotations with higher probabilities. The proposed refinement algorithm adopts progressive method instead of iterative, which can significantly decrease the time cost of refining annotations. In order to further improve efficiency without sacrificing precision, we propose the block-based normalized cut algorithm to segment image. Experiments conducted on standard Washington Ground Truth Image Database demonstrate the effectiveness and efficiency of our proposed approach for refining image annotations.
机译：自动图像标注是缩小低层内容与高层语义概念之间语义鸿沟的一种有前途的解决方案，它已成为图像检索，模式识别和机器学习领域的活跃研究领域。但是，即使是最专用的注释算法也常常不能令人满意。图像标注的细化最近吸引了更多的关注。提出了一种基于互信息的动态投票改进算法。与传统的细化算法不同，该算法采用动态加权投票来度量候选注释之间的相关性，不仅允许概率较高的注释拒绝概率较低的注释，而且还允许概率较低的注释拒绝概率较高的注释。具有较高概率的注释。提出的细化算法采用渐进式方法而不是迭代式方法，可以显着减少细化注释的时间成本。为了在不牺牲精度的情况下进一步提高效率，我们提出了基于块的归一化分割算法对图像进行分割。在标准的华盛顿地面真相图像数据库上进行的实验证明了我们提出的提炼图像注释方法的有效性和效率。

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