摘  要：随着信息数字技术的飞速发展，各行各业的大量信息都在以多媒体信息的方式被数字化。图像是最为广泛和基本的多媒体信息，图像数据的应用领域己涉及到科学技术和日常生活的各个方面。基于内容的图像检索(CBIR, Content-Based Image Retrieval)是一种用图像的视觉特征(颜色、纹理、形状等)进行图像检索的技术。基本思路是根据某种算法提取已给定的图像的特征，再依靠相似性度量函数，把图像库中图像特征与范例图像的特征进行匹配，然后按照相似度髙低排序，根据排序显示检索的结果。在整个检索过程中，特征提取和相似性度量是关键问题，前者将影响到图像内容的描述，而后者则会对计算的实时性提出比较高的要求。本文主要围绕CBIR中图像特征（颜色、纹理、形状等）提取的关键技术展开研究，具有较高的理论意义和实际应用价值。93794

毕业论文关键词：基于内容的图像检索，特征提取，颜色，纹理，形状

Abstract: With the rapid development of information and digital technology, a great deal of information in all walks of life is being digitized in the form of multimedia information。 Image is the most extensive and basic multimedia information, and the application field of image data has involved all aspects of science and technology and daily life。 Content based image retrieval (CBIR, Image, Retrieval, Content-Based) is a technique of image retrieval using visual features (color, texture, shape, etc。) of images。 The basic idea is based on the characteristics of the image extraction algorithm has some given, then rely on the similarity function, the matching features of images and examples of image features, and then according to the similarity of high low ranking, ranking according to display search results。 In the entire retrieval process, feature extraction and similarity measurement are the key issues。 The former will affect the description of image content, while the latter will put forward higher requirements for real-time computing。 This paper mainly focuses on the key technologies of image features (color, texture, shape, etc。) extracted from CBIR, which has high theoretical significance and practical application value。

Keywords: content based image retrieval, feature extraction, color, texture, shape

目   录

1 引言 4

2 图像特征表示与分析 5

2。1 颜色特征 5

2。1。1 颜色模型 6

2。1。2 常用图像的颜色特征提取 8

2。2 纹理特征 10

2。3 形状特征 11

3 聚类算法和相似性度量 13

3。1 模糊聚类算法 13

3。1。1 FCM聚类算法 13

3。1。2 FCM聚类算法的参数 14

3。2 距离相似性度量 15

3。3 归一化 16

4 基于内容的图像检索方法 16

4。1 检索效果评价方法 16

4。2 动态调整多特征权值的检索方法 17

5 实验结果和性能分析 19

总  结 21

参考文献 22

致  谢 23