十种基于颜色特征图像检索算法的比较和分析  

A Comparitive and Analysis Study of Ten color Feature-based Image Retrieval Algorithms

作　　者：刘忠伟[1] 章毓晋[1] 

机构地区：[1]清华大学电子工程系

出　　处：《信号处理》2000年第1期79-84,共6页Journal of Signal Processing

基　　金：国家高技术计划;国家自然科学基金

摘　　要：在基于内容的图像检索中，颜色特征已得到广泛应用。本文对十种利用颜色特征进行图像检索的算法利用同一图像库进行了实验比较．实验结果表明，无论在HSI空间或MTM空间，累加直方留法均优于一般直方留法，对这一点本文还首次结出了严格的理论证明。实验结果还表明，加权距离法比欧氏距离法总体上没有明显改善，MTM空间比HSI空间也没有显出优势，而中心矩法算法简单，检索速度快，通过调整加权系数，检索精度可以接近累加直方留法．本文的实验和分析对选择和优化检索算法有一定的参考价值．In content-based image retrieval,color features are widely used. This paper reports a comparativesstudy of ten algorithms using images of a common database. The experimental result shows that whether in HSIor in MTM color spase the cumulative histogram based algorithm is much better than the traditional histogrambased algorithm, a mathematical prove is aJso trovital for the first time. the results the show that MTM colorspase has no apparent advantage over HSI color space and using weighted distance as the similarity measure isjust a little better than using Euclidean distance. Ih addition, the results show that the algorithms based on Centralmoments are not only simple but also comparable in performane with the algorithms based on coummulativehistogrm by suitably adapting the weighing coefficients. This work would be helpful for selecting and optimizingretrieval algorithms.

关 键 词：图像检索 颜色特征 算法 图像库 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN919.8[自动化与计算机技术—计算机科学与技术]