计算机应用 ›› 2012, Vol. 32 ›› Issue (04): 1101-1103.DOI: 10.3724/SP.J.1087.2012.01101

• 图形图像技术 • 上一篇    下一篇

基于互补特征的纹理图像检索

曲怀敬   

  1. 山东建筑大学 信息与电气工程学院,济南 250101
  • 收稿日期:2011-09-16 修回日期:2011-11-15 发布日期:2012-04-20 出版日期:2012-04-01
  • 通讯作者: 曲怀敬
  • 作者简介:曲怀敬(1965-),男,山东烟台人,副教授,博士,主要研究方向:模式识别、多尺度多方向图像处理。
  • 基金资助:
    国家自然科学基金资助项目

Texture image retrieval based on complementary features

QU Huai-jing   

  1. School of Information and Electric Engineering, Shandong Jianzhu University, Jinan Shandong 250101, China
  • Received:2011-09-16 Revised:2011-11-15 Online:2012-04-20 Published:2012-04-01
  • Contact: QU Huai-jing

摘要: 针对互补特征可以有效地改善图像检索系统性能的特点,提出一种在改进Contourlet变换域采用L1能量与广义高斯分布参数特征的纹理图像检索方法。首先,应用改进的方法对方向子带系数进行广义高斯统计建模。然后,分别单独利用各个特征和相应的相似性测度进行检索。最后,基于直接的相似性测度和,采用这两种互补的特征进行检索。实验结果表明,和采用单一特征相比较,互补特征由于充分地反映了图像的结构信息和随机分布信息,从而有效地提高了纹理图像数据库的平均检索率。

关键词: 改进的Contourlet变换, 建模, L1能量, 广义高斯分布, 互补特征, 纹理图像检索

Abstract: Because the performance of the image retrieval system could be effectively improved by using the complementary features, a retrieval method of the texture image using L1 energy and generalized Gaussian distribution parameter features was proposed in the improved Contourlet transform domain. Firstly, the directional subband coefficients went through generalized Gaussian modeling with an improved approach. Then, the texture images were respectively retrieved based on the single feature and the corresponding similarity measurement. Lastly, using the complementary features and the direct summation of their similarity measurements, the texture images were retrieved. The experimental results show that, compared with single feature, the average retrieval rates of the texture image database are effectively improved by the complementary features that fully represent the structural information and the random distribution information.

Key words: improved Contourlet transform, modeling, L1 energy, generalized Gaussian distribution, complementary feature, texture image retrieval