Semantic-Based Retrieval Using Various Visual Features for Real-World Images

Author(s): 
H. Y. Cui†, J. F. Cao†*, H. Shi‡, & E.C. Bacharoudis§

Affiliation(s): 
Department of Computer Science and Technology, Xinzhou Teachers University, Xinzhou 034000, China,
‡School of Computer Science and Technology, Taiyuan University of Science and Technology,
Taiyuan 030024, China
§Department of Mechanical Engineering, Instituto Superior Técnico, Technical University of Lisbon, Av. Rovisco Pais, 1049- 001 Lisboa, PortugaL

Cite this paper
H. Y. Cui†, J. F. Cao†*, H. Shi‡, & E.C. Bacharoudis§, “Semantic-Based Retrieval Using Various Visual Features for Real-World Images”, Journal of Mechanical Engineering Research and Developments, vol. 39, no. 2, pp. 324-339, 2016.  DOI: 10.7508/jmerd.2016.02.008

ABSTRACT:  Nowadays, the development of multimedia technology has resulted in the rapid growth of digital images and more and more digital images are available. It has become an urgent problem to efficiently retrieve images which can better meet the users’ requirements based on semantic. The description of semantic features for images becomes an issue in multimedia processing. In this paper, we propose novel methods which describe semantic of color, texture and shape features, and propose a method of image retrieval based on three features of color, texture and shape. Firstly, we extract low-level color feature using the method of sub- block, and then apply OCC model for describing the color semantic; Secondly, we use Canny operator to obtain the edge information of image in order to extract low-level texture feature of image, and then applied the improved Tamura model for describing the texture semantic; Thirdly, BP neural network is used to map from low-level features to high-level semantic features; Finally Zernike moment is used to extract shape features and image retrieval is implemented with three features. Choosing Corel image database as testing image database, experiments achieved good effect compared with the method only based on color semantic feature. Experimental results show that the proposed method is capable of meeting the users’ retrieval requirements and can lay a good foundation for solving “semantic gap” problem between low features and high semantic features.

Keywords : Semantic-Based Retrieval; Feature Extraction; Semantic mapping; BP Neural Network; Similarity.

References
[1] Amold W M, Semeulders, et al. Content-Based image retrieval at the end of early years. IEEE Trans. On Pattern Analysis and Machine Intelligence, 22(12) (2000) 1349-1380.
[2] Rui Yong, Huang Thomas S, et al. Image retrieval: past, present, and future. Journal of Visual Communication and Image Representation, 10(1999) 1-23.
[3] Ling He, Lingda Wu, Yichao Cai. Indexing Techniques in CBIR: a Survey. MINI-MICRO SYSTEMS, 27(1) (2006) 141-145.
[4] A. Mojsilovic, B. Rogowitz. Capturing image semantics with low-level descriptors. Proceedings of the ICIP, 9 (2001) 18–21.
[5] X.S. Zhou, T.S. Huang, CBIR: from low-level features to high-level semantics. Proceedings of the SPIE, Image and Video Communication and Processing, 3974 (1) (2000) 426–431.
[6] A.W.M. Smeulders, M. Worring, A. Gupta, R. Jain. Content-based image retrieval at the end of the early years. IEEE Trans. Pattern Anal. Mach. Intell. 22 (12) (2000) 1349–1380.
[7] Liu Y, Zhang D S, Lu G, et al. A Survey of Content-Based Image Retrieval with High-Level Semantics. Pattern Recognition, 40(1) (2007) 262-282.
[8] Liu Y, Zhang D S, Lu G. Region-Based Image Retrieval with High-Level Semantics using Decision Tree Learning. Elsevier Pattern Recognition, 41(8) (2008) 2554-2570.
[9] Chiang C -C, Hung Y -P, Yang H, et.al. Region-based Image Retrieval Using Color-Size Features of Watershed Regions. Elsevier Journal of Vision Communication and Image Representation, 20(3) (2009) 167-177.
[10] Wang Z, Jia K, Liu P. An Effective Web Content-based Image Retrieval Algorithm by Using SIFT Fea-ture. World Congress on Software Engineering(WCSE), (2009) 291-295.
[11] X.X. Chen, R. zhang. A Multi-scale Phase Feature Based Method For Image Retrieval. Journal of Electronics & Information Technology, 31(5) (2009) 1193-1196.
[12] Zhang R, Zhang L, Wang X-J, et al. Multi-feature pLSA for Combining Visual Features in Image Annotation, Proc.of 19th ACM Inter. Conf. on Multi-media(ACM MM), (2011) 1513-1516.
[13] H.W. Hao, F.Y. huang, J. Zhou. Image Retrieval Method Based on ROI and MCS. PR & AI, 21(2) (2008) 240-245.
[14] Nguyen D, Yap G, Liu Y, et al. A Bayesian Approach Integrating Regional and Global Features for Image Semantic Learning. Proc. of Inter. Conf. on Multimedia and Expo,(2009) 546-549.
[15] Yanai K. Web Image Mining Toward Generic Image Recognition. Proc. of 12th Inter. World Wide Web Conference,(2003)1.
[16] S.H. Feng, C.Y. Lang, D. Xu. Combining Graph Learning and Region Saliency Analysis for Content Based Image Retrieval. ACTA ELECTRONICA SINICA, 39(10) (2011) 2288-2294.
[17] Zhao B, Li F-F, Xing E. Large-Scale Category Structure Aware Image Categorization. Proc. of the Neural Information Processing Systems(NIPS),(2011).
[18] Deng J, Satheesh S, Berg A C, et al. Fast and Balanced: Efficient Label Tree Learning for Large Scale Object Recognition. Proc. of the Neural Information Processing Systems(NIPS),(2011).
[19] Deng J, Berg A C, Li K, et al. What Does Classifying More Than 10,000 Image Categories Tell us. Proc. of European Conf. on Computer Vision(EC-CV),Part V,(2010) 71-84.
[20] Bengio S, Weston J, Grangier D. Label Embedding Trees for Large Multi-Class Tasks. Proc. of Neural Information Processing Systems(NIPS),(2010).
[21] Jacob L, Bach F, Vert J-P. Clustered Multi-Task Learning: A Convex Formulation. Proc. of the Neural Information Processing Systems(NIPS),(2008).
[22] Bakker B, Heskes T. Task Clustering and Gating for Bayesian Multitask Learning. Journal of Machine Learning Research, (4) (2003) 83-99.
[23] Boiman O, Shechtman E, Irani M. In Defense of Nearest-Neighbor Based Image Classification. Proc. of Inter. Conf. on Computer Vision and Pattern Recognition(ICVR),(2008)1-8..
[24] Fergus R, Bernal H, Weiss Y, et al. Semantic Label Sharing for Learning with Many Categories. Proc. of European Conf. on Computer Vision(EC-CV),Part I,(2010) 762-765.
[25] Abdolah C, Golshah N, Alfred M. Sketch-based image matching using angular partitioning. IEEE Trans. on Pattern Systems and Humans, 35 (1) (2005) 28-41.
[26] Andrew Ortony, Gerald L. Clore, Allan Collins. The Cognitive Structure of Emotions. Cambridge, UK:Cambridge University Press,(1988).
[27] J.J. Chen, H.F. Li, J.Xiang, J.J. Zhao. Analysis technology of image emotion semantic. Beijing: Electronics industry press, (2011).