基于多特征哈希的图像检索的研究

发布时间：2018-03-29 19:15

本文选题：图像检索　切入点：图像特征　出处：《吉林大学》2017年硕士论文

【摘要】：随着互联网技术的快速发展和数字产品的普及,大量的图像和视频被上传到互联网,并且其内容信息正在以惊人的速度急剧膨胀。如何在这些海量的信息资源中快速检索相关的信息已经成为一个重要的研究问题。本文重点研究基于内容的大规模快速图像检索技术。传统的基于内容的图像检索集中致力于图像的可视化信息的提取,如图像的颜色、纹理和局部关键点特征。但是随着图像数据集的不断增大,这些传统的方法在效率上和实时性上遇到了很大的挑战,而且单一特征往往不能完整的表达一幅图像的内容信息。基于这些现状,本文提出采用多特征融合哈希算法来实现大规模图像的快速检索。本文将图像的全局和局部特征进行融合,与此同时学习出紧凑的哈希函数和哈希码。根据哈希码序列的二值性,采用快速的汉明距离和匹配个数统计计算出两幅图像的相似度,最后将检索结果按照相似度从大到小的顺序展示出来。为了验证本文的算法,我们选取了两个具有代表性的公共数据集。可靠的实验结果表明本论文提出的算法无论在实时性还是高效性等方面都优于其他优秀的算法,并且很好的解决了算法可扩展性的重要问题。
[Abstract]:With the rapid development of Internet technology and the popularity of digital products, a large number of images and videos have been uploaded to the Internet. And its content information is expanding rapidly at an alarming speed. How to quickly retrieve the relevant information in these massive information resources has become an important research problem. This paper focuses on the large-scale research based on content. Traditional content-based image retrieval focuses on the extraction of visual information from images. Such as image color, texture and local key point features. But with the increasing of image data set, these traditional methods face great challenges in efficiency and real-time. And a single feature often doesn't fully represent the content of an image. In this paper, a multi-feature fusion hashing algorithm is proposed to realize the fast retrieval of large scale images. In this paper, the global and local features of the image are fused. At the same time, the compact hash function and the hash code are learned. According to the binary property of the hash code sequence, the similarity between the two images is calculated by using the statistics of the hamming distance and the matching number. Finally, the retrieval results are displayed in the order of similarity from large to small. We select two representative common data sets. The reliable experimental results show that the proposed algorithm is superior to other excellent algorithms in real time and high efficiency. And it solves the important problem of algorithm extensibility.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】