一种基于显著兴趣点的图像检索方法  

Image Retrieval Method Based on Distinct Interest Points

作　　者：程涛[1] 侯榆青[1] 李明俐[1] 常哲[1] 

机构地区：[1]西北大学信息科学与技术学院,西安710127

出　　处：《计算机工程》2010年第18期171-173,共3页Computer Engineering

基　　金：国家自然科学基金资助项目"文物三维模型的语义标注与本体技术检索研究"(60873094);西北大学研究生自主创新基金资助项目(08YZZ34)

摘　　要：提出一种利用显著兴趣点结合颜色矩和距离直方图进行图像检索的方法。该方法将兴趣点作为图像中用户关注的主要视觉线索,包括显著兴趣点检测、距离直方图和颜色矩的特征提取3个步骤,既利用兴趣点的局部颜色特征,又考虑兴趣点间的空间距离关系,克服了传统颜色矩没有空间位置信息的缺陷。实验结果表明,该方法实现简单,能够有效提高图像检索的效率。This paper proposes a new image retrieval nethod which utilizes distinct interest points of the image, and combines color moments and the distance histogram. Interest points of the image are regarded as the main visual cues concerned by users, and it includes detecting distinct interest points, calculating the color distance histogram and color moments. This method not only uses the local color characteristics of an interest point, but also considers space relations of the distance between interest points, and overcomes the traditional color moments＇ lacking location information. Experimental results show that the method is simple, and it can effectively increase the efficiency of the image retrieval.

关 键 词：显著兴趣点 局部颜色特征 距离直方图 颜色矩 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]