一种利用颜色词检索图像的方法  

A Method of Image Retrieval with Color Names

作　　者：孟祥增[1] 王玲[1] 李海霞[1] 钟义信[2] 

机构地区：[1]山东师范大学传播学院,济南250014 [2]北京邮电大学智能中心,北京100867

出　　处：《中国图象图形学报（A辑）》2005年第3期349-353,共5页Journal of Image and Graphics

基　　金：国家自然科学基金项目(69982001)

摘　　要：在基于内容的图像检索中,图像的内容包括图像的低层视觉特征和高层语义。通常,前者是通过图像分析得到的数据,后者是用自然语言表示的概念词,但两者没有明显的联系。为了能够对彩色图像进行更有效的检索,试图将彩色图像的主色调数据与自然语言检索中的颜色词相结合,用于图像检索,并首先通过分析常用的颜色模型,提出了一种新的颜色模型和一种颜色相似度计算方法,然后在此基础上提出了一种提取图像主色调的方法,最后通过分析颜色命名系统,提出了一种利用颜色词检索彩色图像的方法,同时给出了初步实验结果。实验表明,这种方法在彩色图像检索中是有效的。The contents of image include low level visual features and high level semantic in image retrieval based on content. Generally, the former is data obtained from image analysis and the latter is describing concepts in nature language, but they are not related to each other simply. The paper presents a trial with combining the main colors of color image and color names in Chinese language in color image retrieval. A new color model and a method of calculating the similarity between two colors are proposed after a few of color models are analyzed, on which a method of extracting the main colors of an image is given. The color names are discussed and a method of retrieving color images with color names is proposed. The preliminary retrieving results obtained on trial are discussed, which shows that the method is quite helpful in color images retrieval.

关 键 词：彩色图像检索 颜色模型 基于内容的图像检索 相似度计算 数据 视觉特征 表示 主色调 自然语言检索 方法 

分 类 号：TB482[一般工业技术—包装工程] TP391[自动化与计算机技术—计算机应用技术]