基于内容的文物图像检索技术的研究和 实现-计算机应用技术专业论文.docxVIP

万方数据 万方数据 目录 目录 摘 要 我国拥有丰富的文物资源，但是对这些文物资源的保存却十分困难。在“大 数据”的时代背景下，为了更好的对文物资源进行鉴赏、交流和保护，建立一 个完善的数字化文物保护体系势在必行。 作为馆藏文物数字化进程的重要组成部分，文物图像检索技术逐渐成为新 的研究热点。文物图像本身具有数据量大、维度和信息分辨率高等特点，传统 的检索方法已不能实现对海量文物图像的快速精准检索。基于内容的图像检索 技术（CBRI ）应运而生，并迅速成为图像处理和计算机视觉领域的技术新贵。 本文对 CBIR 的关键技术进行探索了与研究，旨在构建一个性能良好的文 物图像检索模型。所做的主要工作有： （一）研究了图像底层视觉特征的提取算法。通过对大量实验数据的对比 分析，提出了综合使用颜色特征和 LBP-HF 纹理特征的多特征融合方法来实现 对文物图像的检索,实验证明该方法的检索准确率高于单特征检索。 （二）探索了图像的相似性度量标准。因为文物图像中颜色信息和纹理信 息在图像检索时的所起到的重要性不同，提出了对不同特征的相似度进行加权 求和的方法，并将其作为图像特征的综合相似性度量标准。 （三）将基于图像底层视觉特征的图像检索技术与相关反馈技术相结合， 避免了因缺乏高层语义而产生的“语义鸿沟”。通过对用户反馈信息的不断学 习，调整检索策略,进而达到提高文物图像检索性能的目的。 关键词： CBIR 特征提取 相似性度量 相关反馈 I Abstract China has rich cultural resources, but the preservation of these cultural relics is very difficult. In order to have a better appreciation and protection of these resources, the establishment of a complete digital heritage conservation system is imperative. As an important part of cultural relics digital process, image retrieval technique has become a new research hotspot. Heritage image itself has the characteristics of large amount of data, high dimensions and resolution. So traditional retrieval methods cannot achieve a fast and accurate retrieval of massive cultural image. Then, Content-based image retrieval (CBRI) came into being, and quickly became the technology upstart in the field of image processing and computer vision. In this paper, the key technologies of CBIR were explored and research. We also built an image retrieval model with good performance. The main work includes: (一) We studied the extraction algorithm of image low-level visual features. Through the comparison and analysis of a large number of experimental data, we developed a comprehensive use of color features and LBP-HF texture features, as a multi-feature fusion to achieve a better image retrieval result. Experiments show that the retrieval accuracy is higher than that of the single feature search. (二) The similarity measure of images has been explor