计算机应用 ›› 2010, Vol. 30 ›› Issue (11): 2986-2987.

• 模式识别 • 上一篇    下一篇

基于高斯混合模型的自动图像标注方法

陈娜   

  1. 石家庄铁路运输学校
  • 收稿日期:2010-04-13 修回日期:2010-06-20 发布日期:2010-11-05 出版日期:2010-11-01
  • 通讯作者: 陈娜

Automatic image annotation method based on Gaussian mixture model

  • Received:2010-04-13 Revised:2010-06-20 Online:2010-11-05 Published:2010-11-01

摘要: 为了进一步完善自动图像标注方法,提出基于高斯混合模型的自动图像标注方法。该方法通过建立每个关键词唯一的高斯混合模型(GMM),准确地描述关键词的语义内容,进而提高自动图像标注的精确性。最后,通过采用COREL图像数据集与不同方法的比较,从平均查准率、平均查全率的实验结果验证了该方法的有效性。

关键词: 高斯混合模型, 自动图像标注, 机器翻译, 语义鸿沟, 聚类

Abstract: Automatic image annotation already becomes a feasible way to reduce "semantic gap". In order to improve the performance of automatic image annotation, Gaussian Mixture Model (GMM) based automatic image annotation method was proposed in this paper. GMM was built for each keyword to accurately characterize its semantic content. Simultaneously, the precise results can be obtained to promote the performance of image annotation. At last, based on the COREL database, the proposed method is verified to be effective in average precision and average recall.

Key words: Gaussian Mixture Model (GMM), automatic image annotation, machine translation, semantic gap, clustering