基于区域和全局特征的图像检索系统设计与实现-计算机技术专业论文.docx

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract With the development of science and technology, and constantly improvement of peoples living standard, the image and video and other multimedia services people can enjoy with become more and more rich. Image retrieval technology has made great development. It can be divided into two kinds. One is to use the underlying characteristics of the image as the global feature for the image retrieval. This method measures the image as a whole, but ignored the local information of image. And image retrieval based on region due to the use of local information of the image, you can retrieve image at the object level to a certain extent, and thus it became a hot research topic in recent years. But it is faced with the shortcomings of region extraction imperfect. By using regional features and global features for image retrieval can avoid to faults of using global features to a certain extent, but also make use of the advantages of image retrieval based on region. On regional feature extraction, using the Grabcut algorithm to get the interested region, then using Gabor texture to describe it, the color histogram is proposed to describe the global features. In image feature matching calculation, usually need to match all of the images, it faces the problem of the large amount of calculation, so can use a secondary retrieval model, clustering for image when matching by using global features for a filter, and then do image matching in the filtered. Experiments show that in image retrieval, regional and global features can be used to achieve good results. Key words：Region-of-interest，Image segmentation，Image retrieval，Secondary retrieval PAGE IV PAGE IV 目 录 摘 要 I Abstract II 1 绪论 1.1 课题的研究背景 (1) 1.2 课题的研究目的和意义 (2) 1.3 国内外研究现状 (3) 1.4 主要研究内容和组织结构 (7) 2 相关技术 2.1 图像特征描述 (9) 2.2 相似性度量 (13) 2.3 图像分割 (15) 2.4 本章小结 (16) 3 系统总体设计 3.1 需求分析与用例分析 (17) 3.2 图像检索系统框架 (19) 3.3 检索流程 (21) 3.4 特征描述和相似性计算 (22) 3.5 本章小结 (26) 4 感兴趣区域提取 4.1 Gra