J4 ›› 2006, Vol. 28 ›› Issue (7): 62-64.
• 论文 • 上一篇 下一篇
胡一君[1] 邹北骥[2]
出版日期:
发布日期:
Online:
Published:
摘要：
图像等分成M×N块后,将子块分成背景子块、目标子块和边缘子块三类,并从中提取颜色、空间特征和边缘特征,求图像间相似度时只在同类子块之间进行匹配.这样既减少了匹配运算量,又可避免不同类子块匹配所产生的干扰.实验结果表明,该方法不仅求图像间相似度的运算量小,而且对图像的旋转和平移变化不敏感,具有较好的检索性能.
关键词: 基于内容的图像检索 分类子块 相似度
Abstract:
After an image is evenly divided into a number of M×N non-overlapped blocks, these blocks are classified into three classes., background blocks, object blocks and edge blocks. The extraction methods for the features are given and a practical CBIR system is designed. Experimental results demonstrate t he efficiency of our techniques.
Key words: content-based image retrieval classified block similarity
胡一君[1] 邹北骥[2]. 基于分类子块的图像检索[J]. J4, 2006, 28(7): 62-64.
0 / / 推荐
导出引用管理器 EndNote|Ris|BibTeX
链接本文: http://joces.nudt.edu.cn/CN/
http://joces.nudt.edu.cn/CN/Y2006/V28/I7/62
