基于兴趣点的图像检索方法研究  

Research of Image Retrieval Method Based on Interest Points

作　　者：孙艳红[1,2] 林鹰[1] 魏海芳[1] 

机构地区：[1]重庆交通大学管理学院,重庆400074 [2]重庆交通大学图书馆

出　　处：《计算机时代》2010年第4期1-3,共3页Computer Era

基　　金：国家自然科学基金项目(60603027);重庆市科委自然科学基金计划项目(2009BB2375)

摘　　要：提出了基于兴趣点的多种图像特征相结合的图像检索新方法。该方法利用兴趣点周围局部区域的环形颜色直方图和Gabor小波变换提取纹理特征作为刻画图像的主要特征,并结合兴趣点的空间分布对图像进行检索。该方法不仅弥补了单一特征无法真正表征图像的缺陷,还保证了检索算法对图像旋转、平移的识别不变性。与同类方法相比,该方法有效提高了图像检索的准确率。A new image retrieval method combining many image features based on interest points is presented. The method extracts texture features as the main characteristics to depict image by using the annular color histogram around the local area of interest points and Gabor wavelet transformation, and retrieves image by combining the space distribution of interest points. It not only overcomes the defect not able to truly characterize image by using single feature, but also ensures the recognition invariance of retrieval algorithm for image rotation and translation. Compared with the similar methods, this method can effectively improve the accuracy rate of image retrieval.

关 键 词：兴趣点 环形颜色直方图 纹理特征 图像检索 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]