基于金字塔多尺度LBP和自学习哈希的面料图像检索算法研究

发布时间：2019-06-18 08:28

【摘要】：随着互联网与多媒体信息技术的快速发展,基于内容的图像检索已经成为了信息领域的热点研究课题,并广泛应用于公安系统、医疗系统、知识产权维护系统等。本文关注图像检索在纺织行业的应用,探究适用于面料图像的检索方法,以满足用户越来越多的时尚需求。本文对面料图像的特性进行了深入分析,从特征提取、索引构建两个方面对检索算法进行了研究,提出基于金字塔多尺度LBP和自学习哈希的面料图像检索算法。本文的主要工作和创新点如下：1.提出了金字塔多尺度LBP特征提取算法。本文介绍了典型的LBP算子,并深入分析了几种现有多尺度LBP算子的不足,在结合面料图像特性的基础上,提出一种金字塔多尺度LBP特征提取算法。该算法首先对面料图像进行金字塔分解,然后对每层金字塔图像提取等采样点的多半径LBP算子。与其它现有的多尺度LBP算法相比,本文提出的金字塔多尺度LBP算法的特征提取范围更广,特征表达能力更强,能够有效地描述面料图像的纹理信息。2.将自学习哈希算法应用于面料图像检索。对几种哈希技术进行了分析,针对现有方法的不足,提出将用于文本信息检索的自学习哈希算法应用到图像检索中,并针对图像与文本数据之间的差异性,对自学习哈希过程进行了改进。利用改进后的自学习哈希算法对面料图像建立索引,大大节约了数据存储空间,减少了计算量。3.在前两种算法的基础上,提出基于金字塔多尺度LBP和自学习哈希的面料图像检索算法。该算法采用金字塔多尺度LBP算子提取面料图像的特征,结合了图像的局部信息与全局信息,有效地描述了面料图像的纹理结构。在索引构建阶段采用改进的自学习哈希算法,有效的保留了原始空间图像数据的相似结构。实验证明,本文提出算法的检索性能普遍优于其它相关的算法。
[Abstract]:With the rapid development of Internet and multimedia information technology, content-based image retrieval has become a hot research topic in the field of information, and is widely used in public security system, medical system, intellectual property maintenance system and so on. This paper focuses on the application of image retrieval in textile industry, and explores the retrieval methods suitable for fabric image in order to meet the more and more fashion needs of users. In this paper, the characteristics of fabric image are deeply analyzed, the retrieval algorithm is studied from two aspects of feature extraction and index construction, and a fabric image retrieval algorithm based on pyramid multi-scale LBP and self-learning hash is proposed. The main work and innovations of this paper are as follows: 1. A pyramid multi-scale LBP feature extraction algorithm is proposed. In this paper, the typical LBP operators are introduced, and the shortcomings of several existing multi-scale LBP operators are deeply analyzed. Based on the characteristics of fabric images, a pyramid multi-scale LBP feature extraction algorithm is proposed. Firstly, the fabric image is decomposed into pyramids, and then the multi-radius LBP operator of sampling points is extracted from each layer of pyramid image. Compared with other existing multi-scale LBP algorithms, the pyramid multi-scale LBP algorithm proposed in this paper has a wider range of feature extraction and stronger feature expression ability, and can effectively describe the texture information of fabric images. 2. The self-learning hash algorithm is applied to fabric image retrieval. In this paper, several hash techniques are analyzed. In view of the shortcomings of the existing methods, the self-learning hash algorithm used in text information retrieval is applied to image retrieval, and the self-learning hash process is improved according to the difference between image and text data. The improved self-learning hash algorithm is used to index the fabric image, which greatly saves the data storage space and reduces the amount of computation. On the basis of the first two algorithms, a fabric image retrieval algorithm based on pyramid multi-scale LBP and self-learning hash is proposed. The algorithm uses pyramid multi-scale LBP operator to extract the features of fabric image, combines the local information and global information of the image, and effectively describes the texture structure of fabric image. In the index construction stage, the improved self-learning hash algorithm is used to effectively retain the similar structure of the original spatial image data. The experimental results show that the retrieval performance of the proposed algorithm is generally better than that of other related algorithms.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】