摘要
现行的基于图像数据挖掘,主要采用聚类分析方法,但聚类分析方法的不足,是图像应该被划为几类受很多因子的影响,很难给出有说服力的证据及证明。本文针对此缺陷,提出了关于图像数据挖掘的一种新思想,即引入小波理论中的Zerotree算法来确定图像初始类的个数及位置,然后再利用聚类算法进行划分。实验表明,用结合了Zerotree的图像数据挖掘的方法,较原图像数据挖掘算法给出合理的解释,提高了聚类效率及图像的相似度.
The current image-based data mining mainly uses the cluster analysis method,but the shortage of the cluster analysis method is how many types the image should be classified,which is affected by many factors influence.And it's difficult to provide convincing evidence and proof.In this paper,the defect is proposed on the image data mining as a new idea,namely,the introduction of wavelet theory in the Zerotree algorithm to determine the image number and location of the initial class,and then divided the use of clustering algorithms.Experimental results show that using a combination of Zerotree image data mining method is given a more reasonable explanation for improved efficiency and image similarity clustering than the original image data mining algorithms.
出处
《微型电脑应用》
2010年第8期26-28,5,共3页
Microcomputer Applications
基金
大连市财政局
大连市信息产业局文件(大财企[2008]281号)基于关联规则算法在物流网站上的应用
