一种新的基于颜色的图像检索算法  

A New Algorithm of Color Based Image Retrieval

作　　者：冯玉才[1] 程珺[1] 聂晶[1] 梁俊杰[1] 

机构地区：[1]华中科技大学计算机科学与技术系,武汉430074

出　　处：《计算机工程与应用》2006年第22期52-55,99,共5页Computer Engineering and Applications

摘　　要：由于图像的底层特征与图像的高层语义特征有较大差异,传统的基于颜色的图像检索效果常常不能尽如人意。为了滤除图像目标检索的干扰因素,突出图像的语义特征并提高图像检索的精度,文中提出了一种新的基于颜色的图像检索算法,该算法将一幅图看作一个事务集,利用了Apriori算法在此事务集中对图像频繁同时出现的颜色进行查找,得到本图像的语义目标相关的行,然后再对这些行进行特征比较。实验结果表明,这种算法的检索效果比传统的图像检索算法更加精确。Due to the semantic gap between low-level feature and high-level semantic feature of an image,people are often not satisfied with the results of traditional color-based image retrieval,For the sake of leaching the disturbing element,emphasizing the semantic feature of the image,and improving the precision,this paper introduces a new algorithm of color-based image retrieval,This algorithm regards an image as a transaction set,uses the Apriori algorithm to find the colors which appear frequently in the same transaction,gets the rows which are correlative with the image＇s semantic object and compares the feature of these correlative rows.The experiment result denotes that the performance of this algorithm is better than traditional image retrieval algorithm.

关 键 词：图像数据库 事务 APRIORI 频繁项集 语义特征 目标行 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]