节点文献
基于纹理特征的图象检索研究
The Research on Texture-Based Image Retrieval
【作者】 丛鑫;
【导师】 孙劲光;
【作者基本信息】 辽宁工程技术大学 , 计算机科学技术, 2007, 硕士
【摘要】 基于内容的图象检索技术是近年来国内外在图象数据库研究中出现的一个新热点问题。本文围绕图象特征提取和图象相似性度量两个问题展开论述。以Gabor基小波提取图象特征理论和线形规划理论为基础,提出用Gauss滤波器进行二次滤波的方法对Gabor小波变换进行改进,从而抑制了噪声,使不同纹理区域内的纹理特征保持稳定,而后通过对纹理特征值的重排序,使其拥有旋转不变性。最后应用单纯形法计算出例子图象与检索图象间的相似性程度数值。本文设计实现了一个基于纹理特征的图象检索系统,对于给定的例子图象,系统搜索指定目录并根据与例子图象的相似性程度数值按升序排列相符合的图象。
【Abstract】 Content Based Image Retrieval is a new hot problem appeared in the local and outside portrait database research in recent years. This text is around the distilling of the character of texture and tolerance of the similarity. With the theory of Gabor filter distilling and the linear programming, this text puts forward a new method that using the Gauss filter to filtrate again. It restrains the yawp and Make the different veins in the district characteristic keep the stability. Then according to compositor of the texture again, it makes its own the constant of revolve. It applied the pure form method computes the example portrait finally and inspect the alike degree number of the portrait.This text designs a portrait for according to veins characteristic index system, settles for the example portrait that give the system searches for to specify the catalogue and according to with the alike degree number of the example portrait presses rise preface arrange agree with the portrait that match.
- 【网络出版投稿人】 辽宁工程技术大学 【网络出版年期】2008年 06期
- 【分类号】TP391.3
- 【被引频次】3
- 【下载频次】145