导　　师： 文灏

学科专业： 0810

授予学位： 硕士

作　　者： ;

机构地区： 华中科技大学

摘　　要： 从20世纪70年代开始，检索技术日益发展，一开始是基于文本的图像检索，后来基于内容的图像检索渐渐地出现。目前，图像检索已经成为当今图像领域研究的热点，而基于内容的检索是图像检索的一个重要分支。基于内容的图像检索本质上是利用图像的视觉特征进行图像特征提取和图像匹配的过程， 本文首先根据发展的背景介绍了基于内容的图像检索的由来，并简明扼要地描述了国内外的CBIR的发展，还说明了图像检索技术目前存在的一些主要问题；接下来探讨了基于内容的图像检索的重点内容，即图像特征的提取，然后用了三个章节分别介绍了基于单一的低层特征检索过程，第一步先选择合适的特征进行特征提取，然后在提取的特征的基本上选择合适的相似度计算，计算出图像的相似度，从而进行图像检索，低层特征主要包括了图像的颜色、纹理和形状三大特征；其次介绍了基于内容的图像检索系统实现方法，主要包括系统的功能描述、总体结构和主要流程。最后对本文的图像检索技术和实现方法进行了总结，指出了一些缺陷。 通过实验并比较结果可得出，基于颜色、纹理和形状中的三者或两者的综合特征的图像检索比基于某一特征检索的效果要好，另外我们可以通过调整三种特征所占权值的大小来得到满意的检索效果。 Since the1970s, the development of fast retrieval technique, slowly retrieved from atext-based image retrieval based on the transition to. Currently, image retrieval hasbecome a hot research field images, and content-based image retrieval retrieval is animportant branch. Content-based image retrieval using the image is essentially visualfeatures for image feature extraction and image matching process, This paper first discusses the content-based image retrieval origin, then introduces thedevelopment of domestic and international current image retrieval, image retrievaltechnology finally shows some existing problems; Then explore the content-based imageretrieval key elements, namely image feature extraction, and then use the three chapterswere introduced low-level features based on a single retrieval techniques and methods,the first low-level features extracted image, and on the low-level features in the imagematching, the image similarity is calculated, thereby performing image retrieval, imagelow-level features including color, texture and shape of the three characteristics; secondlyintroduced content-based image retrieval system implementations, including functionaldescription of the system, the overall structure and main processes. Finally, this imageretrieval technologies and implementation methods are summarized, pointed out somedeficiencies. Through experiments and compare the results can be drawn, based on color, textureand shape of the three or a combination of these characteristics retrieval feature retrievalbased on a ratio better than the other three features we can adjust the size of the shareweights to be satisfied with the search results.

关 键 词： 特征提取 相似度度量 图像匹配 低层特征 图像检索

领　　域： [自动化与计算机技术] [自动化与计算机技术]