| 引用本文: | 钱智明,钟平,王润生.结合非负张量表示与扩展隐Dirichlet分配模型的图像标注.[J].国防科技大学学报,2014,36(6):152-157.[点击复制] |
| QIAN Zhiming,ZHONG Ping,WANG Runsheng.Extended latent Dirichlet allocation for image annotation of nonnegative tensor representation[J].Journal of National University of Defense Technology,2014,36(6):152-157[点击复制] |
|
| |
|
|
| 本文已被：浏览 8950次 下载 6739次 |
| 结合非负张量表示与扩展隐Dirichlet分配模型的图像标注 |
| 钱智明, 钟平, 王润生 |
|
(国防科技大学 电子科学与工程学院， 湖南 长沙 410073)
|
| 摘要: |
| 由于“语义鸿沟”的存在，自动图像标注是一项极具挑战性的工作。考虑到图像低层视觉特征与高层语义概念的差异，分别从图像表示与语义建模两个方面来实现自动图像标注。在图像表示方面，提出了一种正则化约束下的非负张量表示方法，用以提取符合人眼视觉直观理解的图像高阶结构特征。在语义建模方面，提出了一种三层贝叶斯模型——扩展隐Dirichlet分配。该模型利用隐变量来实现图像与标注词的关联，并通过一种基于变分推理的期望最大值方法来估计参数。实验结果表明，ELDA模型在大规模数据库NUS-WIDE上的标注结果相较于现有方法有了显著的提高。 |
| 关键词: 图像标注 非负张量表示 扩展隐Dirichlet分配 变分推理 |
| DOI：10.11887/j.cn.201406027 |
| 投稿日期：2014-03-31 |
| 基金项目:国家自然科学基金资助项目（61271439） |
|
| Extended latent Dirichlet allocation for image annotation of nonnegative tensor representation |
| QIAN Zhiming, ZHONG Ping, WANG Runsheng |
|
(College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| Automatic image annotation is a challenge task due to the well-known semantic gap. Considering the difference between low-level visual features and high-level semantic concepts, the framework of automatic image annotation from the two aspects, image representation and semantic modeling, was constructed. For image representation, a new method of regularized nonnegative tensor representation (RNTP) was presented to abstract the detailed high-order tensor structures according to human’s intuitive recognition. A three-level hierarchical Bayesian model, extended latent Dirichlet allocation (ELDA), was developed for semantic modeling. In ELDA, each item of multiple image factors was modeled as a finite mixture over latent variables. Meanwhile, an efficient expectation-maximization algorithm based on variational inference was proposed for parameter estimation. Extensive experimental results are reported on the NUS-WIDE dataset to validate the effectiveness of our proposed solution to the automatic image annotation problem by comparing with other state-of-the-art methods. |
| Keywords: image annotation nonnegative tensor representation extended latent Dirichlet allocation variational inference |
|
|
