-
摘要: 视觉词典方法（Bag of visual words，BoVW）是当前图像检索领域的主流方法，然而，传统的视觉词典方法存在计算量大、词典区分性不强以及抗干扰能力差等问题，难以适应大数据环境.针对这些问题，本文提出了一种基于视觉词典优化和查询扩展的图像检索方法.首先，利用基于密度的聚类方法对SIFT特征进行聚类生成视觉词典，提高视觉词典的生成效率和质量；然后，通过卡方模型分析视觉单词与图像目标的相关性，去除不包含目标信息的视觉单词，增强视觉词典的分辨能力；最后，采用基于图结构的查询扩展方法对初始检索结果进行重排序.在Oxford5K和Paris6K图像集上的实验结果表明，新方法在一定程度上提高了视觉词典的质量和语义分辨能力，性能优于当前主流方法.Abstract: The most popular approach in image retrieval is based on the bag of visual-words (BoVW) model. However, there are several fundamental problems that restrict the performance of this method, such as low time efficiency, weak discrimination of visual words and less robustness. So, an image retrieval method with enhanced visual dictionary and query expansion is proposed. Firstly, clustering by fast search and finding density peaks are used to generate a group of visual words. Secondly, non-information words in the dictionary are eliminated by Chi-square model to improve the distinguishing ability of the visual dictionary. Finally, an efficient graph-based visual reranking method is introduced to refine the initial search results. Experimental results of Oxford5K and Paris6K datasets indicate that the expression ability of visual dictionary is effectively improved and the method is superior to the state-of-the-art image retrieval methods in performance.1) 本文责任编委 刘跃虎
-
图 1 基于视觉词典优化和查询扩展的图像检索方法流程
Fig. 1 The flow chart of image retrieval based on enhanced visual dictionary and query expansion
图 6 在Oxford5K和Oxford5K+Paris6K数据库上的图像检索MAP值
Fig. 6 The MAP of different methods for Oxford5K and Oxford5K+Paris6K database
图 7 EVD+GBQE方法在Oxford5K+Paris6K数据库上的检索结果
Fig. 7 The image retrieval results of EVD+GBQE for Oxford5K+Paris6K database
表 1 视觉单词$w$与各目标类别统计关系
Table 1 Relation between $w$ and categories of each objective
$C_1$ $C_2$ $\cdots$ $C_m$ Total 包含$w_i$的图像数目 $n_{11}$ $ n_{12}$ $\cdots$ $n_{1m}$ $n_{{\rm{1 + }}}$ 不包含$w_i$的图像数目 $n_{21}$ $n_{22}$ $\cdots$ $n_{2m}$ $n_{{\rm{2 + }}}$ Total $n_{{\rm{ + }}1}$ $n_{{\rm{ +}}2}$ $\cdots$ $n_{{\rm{ + }}m}$ $n_{{{m + }}}$ 
下载: 导出CSV
表 2 不同查询扩展方法的图像检索MAP值对比(%)
Table 2 The image retrieval results of different query expansion methods for Oxford5K database (%)
Initial AQE KNNR DQE GBQE All Souls 71.4 79.3 81.8 81.4 83.6 Ashmolean 76.5 81.2 83.1 85.1 87.4 Balliol 73.8 78.4 79.3 80.6 82.5 Bodleian 67.2 70.5 73.4 74.5 74.8 Christ_Church 74.1 78.3 81.5 82.4 83.2 Cornmarket 77.4 82.1 81.8 83.2 84.3 Hertford 85.7 89.2 90.9 91.6 93.2 Keble 86.5 91.6 92.2 93.8 94.4 Magdalen 54.6 61.6 63.8 62.9 63.7 Pitt Rivers 92.4 95.6 95.3 95.1 97.6 Radcliffe cam 74.4 80.8 82.6 84.7 86.1 Average 75.82 80.78 82.34 83.21 84.62
下载: 导出CSV
-
[1] Chen Y Z, Dick A, Li X, Van Den Hengel A. Spatially aware feature selection and weighting for object retrieval. Image and Vision Computing, 2013, 31(12):935-948 doi: 10.1016/j.imavis.2013.09.005 [2] Wang J J Y, Bensmail H, Gao X. Joint learning and weighting of visual vocabulary for bag-of-feature based tissue classification. Pattern Recognition, 2013, 46(12):3249-3255 doi: 10.1016/j.patcog.2013.05.001 [3] Cao Y, Wang C H, Li Z W, Zhang L Q, Zhang L. Spatial-bag-of-features. In:Proceedings of the 2010 IEEE Conference on Computer Vision and Pattern Recognition. San Francisco, CA, USA:IEEE, 2010. 3352-3359 [4] Philbin J, Chum O, Isard M, Sivic J, Zisserman A. Object retrieval with large vocabularies and fast spatial matching. In:Proceedings of the 2007 IEEE Conference on Computer Vision and Pattern Recognition. Minneapolis, USA:IEEE, 2007. 1-8 [5] Nister D, Stewenius H. Scalable recognition with a vocabulary tree. In:Proceedings of the 2006 IEEE Conference on Computer Vision and Pattern Recognition. New York, USA:IEEE, 2006. 2161-2168 [6] Goes J, Zhang T, Arora R, Lerman G. Robust stochastic principal component analysis. In:Proceedings of the 17th International Conference on Artificial Intelligence and Statistics. Reykjavik, Iceland:JMLR, 2014. 266-274 [7] Goswami A K, Jain R, Tripathi P. Automatic segmentation of satellite image using self organizing feature map (SOFM) an artificial neural network (ANN) approach. International Journal of Advanced Research in Computer Science, 2014, 5(8):92-97 http://connection.ebscohost.com/c/articles/100182789/automatic-segmentation-satellite-image-using-self-organizing-feature-map-sofm-artificial-neural-network-ann-approach [8] McLachlan G, Krishnan T. The EM Algorithm and Extensions (Second Edition). Hoboken, New Jersey:John Wiley & Sons, 2008. [9] Sivic J, Zisserman A. Video Google:a text retrieval approach to object matching in videos. In:Proceedings of the 9th IEEE International Conference on Computer Vision. Nice, France:IEEE, 2003. 1470-1477 [10] Yuan J S, Wu Y, Yang M. Discovery of collocation patterns:from visual words to visual phrases. In:Proceedings of the 2007 IEEE Conference on Computer Vision and Pattern Recognition. Minneapolis, USA:IEEE, 2007. 1-8 [11] Fulkerson B, Vedaldi A, Soatto S. Localizing objects with smart dictionaries. In:Proceedings of the 10th European Conference on Computer Vision. Berlin, Heidelberg, Germany:Springer, 2008. 179-192 [12] Perd'och M, Chum O, Matas J. Efficient representation of local geometry for large scale object retrieval. In:Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition. Miami, USA:IEEE, 2009. 9-16 [13] Shen X H, Lin Z, Brandt J, Avidan S, Wu Y. Object retrieval and localization with spatially-constrained similarity measure and k-nn re-ranking. In:Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence, USA:IEEE, 2012. 3013-3020 [14] Chum O, Philbin J, Sivic J, Isard M, Zisserman A. Total recall:automatic query expansion with a generative feature model for object retrieval. In:Proceedings of the 11th IEEE International Conference on Computer Vision. Rio de Janeiro, Brazil:IEEE, 2007. 1-8 [15] Rodriguez A, Laio A. Clustering by fast search and find of density peaks. Science, 2014, 344(6191):1492-1496 doi: 10.1126/science.1242072 [16] Kesom K, Poslad S. An enhanced bag-of-visual word vector space model to represent visual content in athletics images. IEEE Transactions on Multimedia, 2012, 14(1):211-222 doi: 10.1109/TMM.2011.2170665 [17] Zhang S T, Yang M, Cour T, Yu K, Metaxas D N. Query specific rank fusion for image retrieval. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(4):803-815 doi: 10.1109/TPAMI.2014.2346201 [18] Philbin J, Arandjelović R, Zisserman A. Oxford5K dataset[Online], available:http://www.robots.ox.ac.uk/~vgg/data/oxbuildings/, December, 2015. [19] Philbin J, Zisserman A. Paris6K database[Online], available:http://www.robots.ox.ac.uk/~vgg/data/parisbuil-dings/, December, 2015. [20] Arandjelović R, Zisserman A. Three things everyone should know to improve object retrieval. In:Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence, USA:IEEE, 2012. 2911-2918 [21] Xie H T, Zhang Y D, Tan J L, Guo L, Li J T. Contextual query expansion for image retrieval. IEEE Transactions on Multimedia, 2014, 16(4):1104-1114 doi: 10.1109/TMM.2014.2305909 [22] Gao Y, Shi M J, Tao D C, Xu C. Database saliency for fast image retrieval. IEEE Transactions on Multimedia, 2015, 17(3):359-369 doi: 10.1109/TMM.2015.2389616 -
图(7) / 表(2)
计量
- 文章访问数: 1945
- HTML全文浏览量: 399
- PDF下载量: 669
- 被引次数: 0
