通过抽取的特征进行图象检索的算法测试平台

A Test-Bed for Retrieving Images with Extracted Features

基于内容的图象检索近年来得到了广泛的研究 ,人们已提出了许多基于特征的图象检索算法 ,但如何管理、比较、评价、组合应用这些检索算法已成为继续深入研究必须要解决的一个问题 .为了解决此问题 ,建立了一个通过抽取的特征进行图象检索的算法实验平台 .该平台既具有管理功能 (包括管理各种算法、图象库和图片 ) ,又能集成各种算法 ,以综合实现不同的检索功能 (包括递进检索和综合检索 ) .实验结果表明 ,借助平台对算法和图象进行集中管理 ,既可以方便地对各种基于特征的图象检索算法进行比较和评价 。

Content based image retrieval is a hot research subject and has been studied widely in recent years. A number of feature based image retrieval algorithms, such as using color feature, texture feature, shape feature and position information etc., have been proposed in the literature. How to manage, compare, evaluate, and combine these algorithms becomes a challenge problem for further research and improvement in feature based image retrieval. This paper describes a specially constructed test bed for algorithms of retrieving images with extracted features. This test bed has both the functions of management, such as algorithm management, image database management, and individual image management, as well as the functions of integrating different feature based algorithms together for performing image retrieval task. On the basis of algorithm gathering, different retrieval stratagems, such as step up retrieval (including average and cumulative retrieval) and combined retrieval (using more than one independent feature based technique) have been realized. Experiments shows that with the help of this test bed, especially for its centralized control and management of algorithms and images, it is ease not only to compare and evaluate different feature based image retrieval techniques, but also to combine their advantages and to develop new algorithms.