基于人工免疫的图像聚类算法的研究

Research of image clustering algorithm based on artificial immune

随着图像检索系统的发展,合理组织和管理图像数据库已渐渐成为用户检索的关键所在。因此提出了一种基于人工免疫的图像聚类算法,通过模拟抗体捕获抗原的机制,对由Internet上的60幅图像组成的图像库进行了聚类分析。聚类之前提取了图像的平均颜色特征,获得了每幅图像的mean值,同时,还比较了传统K均值聚类算法与本文算法的性能。实验结果表明,算法计算时间少、聚类误差低、聚类能力强,能有效地提高了检索效率。

With the development of image retrieval system,the organizing and managing of image database effectively has been a key of users＇ retrieving.The algorithm of image clustering based on artificial immune has been proposed in this paper,and the image database including of 60 images in the internet has been clustered by simulating the principle of antibody capturing antigen.Before clustering,the average color feature of images has been extracted,and the mean value has been obtained,at the same time,the performance of this algorithm and the K-average clustering algorithm have been compared.The result of experiment has indicated that the algorithm has some merits such as low time complexity,low clustering error,better clustering ability,and can improve the retrieval efficiency.