基于形状特征的图像检索系统的研究  

Research of Image Retrieval System Based on Shape Feature

作　　者：孙爽滋[1] 谷欣超[1] 杨勇[1] 苗静[1] 

机构地区：[1]长春理工大学计算机科学技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2012年第4期174-177,共4页Journal of Changchun University of Science and Technology(Natural Science Edition)

摘　　要：本文对图像检索系统中的形状特征的提取及描述和相似性测度进行了深入研究。以枪械图库为研究对象,在特征提取及描述中,利用最大方差分割技术得到目标区域,用Hu不变矩描述区域。实验结果表明,最大方差法对枪械图像达到了较好的分割效果,有利于后续的图像处理;加权欧氏距离减小了高阶矩幅值波动大对检索结果的影响。This paper deeply researched on shape feature extraction and depiction for image retrieval system,along with comparability estimation.Taking image for gun as object to research,this paper used Otsu algorithm to find objective area,and depicted objective area by Hu invariant moment.Indicated by experimental result,Otsu algorithm made a preferable effect on image segmentation,Euclidean distance with weight reduced the effect of High-order moments to retrieval result.

关 键 词：图像检索 形状特征 图像分割 不变距 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]