一种基于聚类的图像检索方法  

Image Retrieval Method Based on Clustering

作　　者：常小红[1] 卢虹冰[1] 焦纯[1] 张国鹏[1] 见伟平[1] 史纲[1] 乔佳[1] 

机构地区：[1]第四军医大学计算机应用教研室,西安710032

出　　处：《医疗卫生装备》2013年第8期4-6,共3页Chinese Medical Equipment Journal

基　　金：国家自然科学基金(81071220)

摘　　要：目的:改善基于内容的图像检索精度不高的现状,研究一种提高图像检索精度的方法。方法:采用HSV颜色空间模糊量化方法提取图像的颜色直方图特征,在此基础上使用两阶段聚类方法对图像库中图像的特征形成聚类,检索阶段通过比较查询图像特征与每个聚类的中心来返回相关图像。结果:该方法提高了查询图像的查全率、查准率和相关图像在检索结果中的排名。结论:该图像检索方法是合理有效的,检索结果比较令人满意。Objective To enhance the accuracy of the content-based image retrieval. The color histogram feature was extracted as retrieval feature by using fuzzy quantization method in HSV color space. On this basis, the features of images in the image base formed many clusters by two-stage cluster method. In the image retrieval stage, the relevant images were returned by comparing the features between query image and the centre of each cluster. Experimental results showed that this method could improve the recall and precision of query image, and improved the rank or the relevant image in the retrieval results. Conclusion This image retrieval method is reasonable and effective, and its retrieval results are satisfactory.

关 键 词：图像检索 HSV颜色空间 模糊量化 聚类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]