基于颜色特征的图像检索算法研究

摘 要


本文以基于颜色特征的图像检索实例系统为基础，研究RGB与 HSI 颜色空间内，全局直方图，累积直方图，局部累加直方图对图像的描述。以之为特征，再引入欧氏距离的相似性度量方法实现基于内容的图像检索。根据实验结果比较，总结颜色空间、特征描述方法对检索系统性能的影响，从而得出不同算法 所适合的应用范围。

论文第一章介绍了本课题的研究背景和问题概要解决方案。第二章通过对基于颜色特征的图像检索技术的深入分析，得出实例系统使用的核心方法论，包括RGB与HSI 空间的转换，三种直方图实现公式，和欧式距离计算方法等。第三章基于VC++ .net 技术，阐述了图像检索系统实现的设计思路与执行方法，重点介绍之前提出的核心方法论在系统中的实现算法 。第四章对不同算法 条件下的实验数据进行分析比较，总结不同颜色空间或特征描述方法对检索系统效率的影响。第五章分析设计工作中的问题，提出在进一步的工作中，结合web搜索的相关技术，进行了相关讨论。


关键词：基于内容的图像检索，颜色特征，直方图，相似度


Research on Algorithms for Color-based
Image Retrieval


ABSTRACT


This dissertation is chiefly contributed to the CBIR study of color specification. Based on practical system experiment, under RGB and HSI color space, we do research on the efficiency of global color histogram, global accumulated color histogram, and partial accumulated color histogram, which describes the distribution of colors in each image. By this means, color histogram dissimilarity is brought in to achieve high efficiency in color histogram query process.

Thus, a research background and essential solutions is introduced in this topic. Then by further analysis on color-based image retrieval, we make out the key methodology, including: the transfer of RGB and HSI space, formulas to three kinds of color histogram and similarity distance solutions, to fulfill in instance system. Thirdly, from the view of VC++.net development, we put emphasis on how those key methodology is carried out. In Chapter 4 we make analysis from results of system experiment, and summarize the influences by different means. Finally, vision work is put forward.

Key Words:  CBIR, color specification, color histogram, similarity distance