摘要
利用图像全局特征的检索不能很好地检索用户想要的对象内容,而基于分割后各区域特征的检索又过分依赖于复杂的图像分割算法。针对上述两者的缺点,文中提出了一种基于用户感兴趣区域的图像检索算法。该算法首先对样例进行多分辨率树状分解,再由用户选择分解后的任意多个感兴趣的子图,提取子图的特征以进行相似性度量,并应用相关反馈以更好地捕获用户的检索意图。该方法无需对图像进行复杂的分割就能提取对象特征,且经由实验证明具有较高的查全率。
Image retrieval using the feature of the entire image neglects the objects in which the user is interested,while the result of segmented regions- based retrieval excessivdy depends on complicated segmentation algorithm. An image retrieval algorithm based on region -of-interest is proposed in this paper aiming at covering those two shortages. This algorithm firstly decomposes the sample image into sub- images by multiresolution tree,and then the feature of the sub - images of interest selected by the user is extracted for similarity measure. We also investigate the use of feedback to better capture the user' s intention. This approach can extract the feature of object without segmentation, and our experiments show that it can achieve high rate of recall.
出处
《计算机技术与发展》
2006年第3期104-106,109,共4页
Computer Technology and Development
关键词
基于内容的图像检索
感兴趣区域
相关反馈
content - based image retrieval
region - of - interest
relevance feedback