基于图像主色特征的CBIR
Content-Based Image Retrieval By Dominant Color Feature
-
摘要: CBIR(Content Based Image Retrieval)是当前多媒体检索的热点.颜色特征是图像检索的重要特征.为了提高颜色特征的检索效果,在提取图像主色特征的基础上,进一步提取了相应的主色空间分布信息--主色矩特征,并建立了图像主色特征的描述模型.在改进加权二次型相似性度量方法的基础上,提出主色多特征相似性度量方法,并对几种不同的相似性度量方案进行了对比,其中DCME方法在WWW发布方式的CBIR原型系统上取得了较好的实验结果.Abstract: Color is a kind of significant features in image retrieval. A new kind of image feature, namely Dominant Color Moment (DCM), which reflects spatial information of Dominant Color (DC), is taken into account. A feature model of DC features is also established. Several methods of similarity measurement based on the improvement of quadric form distance function are presented and compared. One of these methods, DCME, shows good effect in the CBIR experimental system implemented by means of WWW distribution.
下载: