文档详情

基于特征点空间关系的图像检索技术研究-计算机软件与理论专业论文.docx

发布:2019-03-27约5.3万字共59页下载文档
文本预览下载声明
摘 摘 要 I I II II 本论文中基于特征点空间关系的图像检索算法是基于内容的图像检索算法的 一个重要分支,是对图像底层特征进行深入研究而提出的一种图像检索算法。图 像底层特征的提取和描述是基于内容的图像检索技术的第一步,也是关键技术之 一,特征提取与描述方法好坏直接关系图像检索的性能优劣。此外,鉴于基于内 容的图像检索的应用前景和商用价值,本课题具有研究的意义和必要。 本论文始终围绕如何充分考虑特征点空间关系来进行图像描述,回顾了图像 检索技术的发展以及国内外研究现状和基于特征点空间关系描述方法的研究现 状,针对当前基于特征点空间关系的图像检索技术存在的不足,进行了具体分析 和研究,并提出了两种图像检索算法。 第一方法是,针对环形直方图不能充分考虑特征点空间的缺点提出了先利用 环形直方图法对图像进行粗匹配,然后再利用 Delaunay 三角网精确描述特征点空 间关系,从而达于精确匹配的目的,完成图像检索的方法,并将其应用到商标图 像检索算法中。 第二方法是,针对基于特征点空间关系的商标图像检索算法中 DT 网描述过于 精细而不适用于较为复杂的图像的问题,提出了基于 Hu 不变矩和改进的环形直方 图的图像检索算法,利用 GIS 中的标准差椭圆来描述图像的特征空间分布,使基 于特征点空间关系的图像检索算法通用性更好,使得常规图像都能用该方法检索。 关键词:图像检索,特征点,空间关系,环形直方图,不变矩 Abstract This paper, image retrieval algorithm based on the spatial relations of feature points, which is based on the image low-level characteristics on research, is an important branch of the content-based image retrieval algorithm. The low-level of the image feature extraction and description is seen as the first step in content-based image retrieval technology, which is also been regarded as one of the key technologies, and whether the feature extraction and description method is good or not is closely related to the performance of retrieving images quality. In addition, in view of content-based image retrieval application prospect and commercial value, it is significance and necessary to do some research on this topic. In this paper, we put our focus on how to take the spatial relationship of feature points into full consideration for the purpose of good image description. Before this, the development of image retrieval technology is reviewed, and the domestic and foreign research present situation is introduced and so is the feature points based on spatial relationship description method. Armed to solve the weaknesses of image retrieval technology which is based on spatial relationship of feature points, the concrete analysis and development trend research is well done, and then two kinds of image retrieval algorithm are put forward. The first algorithm is t
显示全部
相似文档