基于内容的图像特征提取算法研究以及实现
发布时间：2019-05-16 08:44
【摘要】：随着信息社会的发展，图像的使用己经渗透到社会的各行各业，日益增多的图像来源为人们提供了丰富的信息。如何快速地搜索有用的图像己变得越来越迫切。当前流行的网络搜索引擎大多基于文本，对于基于内容的图像检索的研究才刚刚起步。将数字图像处理、数据库和图像检索技术结合起来，建立高速、便捷的图像搜索引擎具有重要的理论和应用价值。 基于内容的图像检索技术包括如下几个内容:图像底层特征提取、相似度匹配、索引机制。本文围绕图像底层特征提取展开了研究，将提取颜色特征、纹理特征、形状特征的经典算法进行了融合和改进，并提出了图像自动分割算法，脉冲耦合神经网络和形态学相结合，在图像对象上的标注。在比较分析现有的图像底层特征的特点的基础上,提出了基于颜色主色调和形状特征的综合特征检索方法。通过上述特征提取算法可以方便地提取出图像多方面的特征，，大大增加了图像检索的精度。 本文以Visual Studio2010作为开发工具，Matlab7.0作为实验工具，设计了一个简单的基于内容的图像检索系统。本文设计的CBIR系统基于上述算法实现了图像特征提取，通过允许用户输入参数的方法支持用户的个性化搜索。最后对本系统进行了实验，达到了预期的效果。
[Abstract]:With the development of information society, the use of images has penetrated into various industries of society, and more image sources provide people with a wealth of information. How to search for useful images quickly has become more and more urgent. At present, most of the popular web search engines are based on text, and the research on content-based image retrieval has just started. It is of great theoretical and application value to combine digital image processing, database and image retrieval technology to establish a high speed and convenient image search engine. Content-based image retrieval technology includes the following contents: image bottom feature extraction, similarity matching, index mechanism. In this paper, the extraction of color features, texture features and shape features is studied, and the classical algorithm of extracting color features, texture features and shape features is merged and improved, and an automatic image segmentation algorithm is proposed. Pulse coupled neural network and morphology are combined to mark the image object. On the basis of comparing and analyzing the characteristics of the existing image bottom features, a comprehensive feature retrieval method based on color main tone and shape features is proposed. Through the above feature extraction algorithm, many features of the image can be easily extracted, which greatly increases the accuracy of image retrieval. In this paper, a simple content-based image retrieval system is designed by using Visual Studio2010 as the development tool and Matlab7.0 as the experimental tool. The CBIR system designed in this paper realizes image feature extraction based on the above algorithm, and supports user personalized search by allowing users to input parameters. Finally, the experiment of the system is carried out, and the expected effect is achieved.
【学位授予单位】：湖北民族学院
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41
本文编号：2478163
[Abstract]:With the development of information society, the use of images has penetrated into various industries of society, and more image sources provide people with a wealth of information. How to search for useful images quickly has become more and more urgent. At present, most of the popular web search engines are based on text, and the research on content-based image retrieval has just started. It is of great theoretical and application value to combine digital image processing, database and image retrieval technology to establish a high speed and convenient image search engine. Content-based image retrieval technology includes the following contents: image bottom feature extraction, similarity matching, index mechanism. In this paper, the extraction of color features, texture features and shape features is studied, and the classical algorithm of extracting color features, texture features and shape features is merged and improved, and an automatic image segmentation algorithm is proposed. Pulse coupled neural network and morphology are combined to mark the image object. On the basis of comparing and analyzing the characteristics of the existing image bottom features, a comprehensive feature retrieval method based on color main tone and shape features is proposed. Through the above feature extraction algorithm, many features of the image can be easily extracted, which greatly increases the accuracy of image retrieval. In this paper, a simple content-based image retrieval system is designed by using Visual Studio2010 as the development tool and Matlab7.0 as the experimental tool. The CBIR system designed in this paper realizes image feature extraction based on the above algorithm, and supports user personalized search by allowing users to input parameters. Finally, the experiment of the system is carried out, and the expected effect is achieved.
【学位授予单位】：湖北民族学院
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41
【参考文献】
相关期刊论文 前4条
1 黄祥林,沈兰荪;基于内容的图像检索技术研究[J];电子学报;2002年07期
2 李向阳,庄越挺,潘云鹤;基于内容的图像检索技术与系统[J];计算机研究与发展;2001年03期
3 金韬,任秀丽;图像检索中颜色特征的提取与匹配[J];计算机辅助设计与图形学学报;2000年06期
4 丁险峰,吴洪,张宏江,马颂德;形状匹配综述[J];自动化学报;2001年05期
相关博士学位论文 前3条
1 孙君顶;基于内容的图像检索技术研究[D];西安电子科技大学;2005年
2 崔江涛;高维索引技术中向量近似方法研究[D];西安电子科技大学;2005年
3 赵珊;基于内容的图像检索关键技术研究[D];西安电子科技大学;2007年
本文编号：2478163
本文链接：https://www.wllwen.com/kejilunwen/sousuoyinqinglunwen/2478163.html
