基于最长公共视觉词串的图像检索方法  

Image retrieval with longest common visual substring

作　　者：苗军 崔嵩 段立娟 张璇 许少武 MIAO Jun;CUI Song;DUAN Lijuan;ZHANG Xuan;XU Shaowu(Beijing Key Laboratory of Internet Culture and Digital Dissemination Research,School of Computer Science,Beijing Information Science and Technology University,Beijing 100101,China;Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京信息科技大学计算机学院网络文化与数字传播北京市重点实验室,北京100101 [2]北京工业大学信息学部,北京100124

出　　处：《计算机工程与应用》2018年第15期192-196,234,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61650201);北京市自然科学基金(No.4162058;No.4152005)

摘　　要：词袋模型是图像检索中的一种关键技术。词袋模型中每张图像表示为视觉词在码本中的频率直方图。这样的检索方式忽视了视觉词间对于图像表示很重要的空间信息。提出一种全新的基于最长公共视觉词串的图像检索方法。词串的提取基于视觉词间的拓扑关系,包含很多图像的空间信息。在Holiday数据集上的实验结果表明提出的方法提升了词袋模型的检索效果。The model of Bag-of-Features(BoF)has been widely used in image retrieval.In the BoF model,each image is represented as a frequency histogram of visual words in a codebook.It ignores the spatial context among the visual words,which is important for image representation.This paper proposes a novel image retrieval method based on the longest common visual substring.The substring is extracted based on the topology between the visual words,which contains lots of image spatial information.The experimental results on Holiday image database demonstrate that the proposed method for image retrieval outperforms the original BoF model and its improved Eo-BoW model.

关 键 词：图像检索 词袋模型 最长公共视觉词串 

分 类 号：TN911.73[电子电信—通信与信息系统]