【作者】 孙振权； 肖诗斌； 吕学强； 施水才；

【Author】 Sun Zhenquan~1,Xiao Shibin~(1,2),Lv Xueqiang~(1,2),Shi Shuicai~(1,2) (1 Beijing Information Sciance and Technology Chinese character Information deal centre; 2 Beijing Tuoersi information and technology Co.lim.)

【机构】 北京信息科技大学中文信息处理研究中心； 北京拓尔思信息技术股份有限公司；

【摘要】 传统的基于内容的图像检索主要依赖颜色、纹理、形状、空间关系等单一视觉特征,检索效果往往不够理想。本文对多特征图像检索问题进行了一定的探讨与研究,改进了单一特征距离归一化的方法,并根据归一化的距离和各个特征的权重计算出两图像的综合距离,作为两图像相似性度量依据,从而实现了较好的图像检索效果。更多还原

【Abstract】 Traditionally image retrieval mainly relies on single feature such as color, texture,shape and spatial relationship,etc.,so the result is usually not so good.In this paper, the fusion of multi-feature similarity distance are studied and discussed and an improved method on normalization of similarity distance is put forward.The Integrated distance of the two images we calculate adding the weight of the various features is the basis of similarity measure in two images.In this way,we can reach better retrieval result.更多还原