Biased Discriminant Subspace Learning for Content Based Image Retrieval School of Electrical & Electronic Engineering Supervisor : Assoc Prof. Wang Lipo School of Electrical &Electronic Engineering Co-Supervisor: Assoc Prof. Lin Weisi School of Computer Engineering PhD Student: Zhang Lining
36
Embed
Biased Discriminant Subspace Learning for Content Based ...imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/... · Two-class SVM for CBIR RF One Class for CBIR RF Active SVM for CBIR
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Biased Discriminant Subspace Learning
for Content Based Image Retrieval
School of Electrical & Electronic Engineering
Supervisor : Assoc Prof. Wang Lipo School of Electrical &Electronic
Engineering
Co-Supervisor: Assoc Prof. Lin Weisi School of Computer Engineering
PhD Student: Zhang Lining
Outline
• Background
• Research Object
• Existing Work
• Proposed Method
• Conclusion and Future work
Background
• Rapid growth of the number of images records and
explosive increase of on-line images
How to retrieval the
image you want?!
Background
• Text Based Image Retrieval– Google Search, Bing Search, Baidu Search
• However, text based image retrieval is enough?– Hard to describe the image using words
– Exactly search the query image in database
• Actually, a picture is worth thousands of words!!!
Background
• However, the severe challenge in image retrieval is the “semantic
gap” issue
• Dynamic Interpretation
Similar
Features
Different
Concept
Different
Features
Similar
Concept
Skee Sunset
Research Object
• Extend existing techniques
• Design new technique and try to narrow the “semantic gap” in CBIR
• Devise novel and efficient features and try to narrow the “semantic gap” in CBIR
• Devise a more reasonable and efficient framework for CBIR
Research Problem
• What is query by example?
User
Query Image
Image Database
Final Results
Key techniques in CBIR
• Image feature extraction
– Global features: color, texture ,shape…
– Local features: bag of features, SIFT, Gabor Functions…
– Fisher’s criterion based for discriminant subspace learning
– Manifold learning for preserving the local geometry
Support Vector Machines for CBIR
Two-class SVM for CBIR RF One Class for CBIR RF Active SVM for CBIR RF
Existing Methods
• Support Vector Machines for CBIR• P. Hong, Q. Tian, and T.S. Huang, “Incorporate Support Vector Machines to Content-Based
Image Retrieval with Relevant Feedback,” Proc. IEEE Int’l Conf. Image Processing, pp. 750-753, 2000.
• Y. Chen, X.-S. Zhou, and T.-S. Huang, “One-class SVM for learning in image retrieval,” in Proc. IEEE Int. Conf. Image Processing, 2001
• G. Guo, A.K. Jain, W. Ma, and H. Zhang, “Learning Similarity Measure for Natural Image Retrieval with Relevance Feedback,”IEEE Trans. Neural Networks, vol. 12, no. 4, pp. 811-820, 2002.
• C.-H. Hoi, C.-H. Chan, K. Huang, M. R. Lyu, and I. King, “Biased support vector machine for relevance feedback in image retrieval,” presented at the Int. Joint Conf. Neural Networks, 2004.
• D. Tao, X. Tang, X. Li, and X. Wu, “Asymmetric bagging and random subspace for support vector machines-based relevance feedback in image retrieval,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 28, no.7, pp. 1088–1099, Jul. 2007
Existing Methods
• Feature reweighting• Y. Rui and T.-S. Huang, “Optimizing learning in image retrieval,” presented at the IEEE
Int. Conf. Computer Vision and Pattern Recognition,2000.
• M. L. Kherfi and D. Ziou, “Relevance feedback for CBIR: A new approach based onprobabilistic feature weighting with positive and negative examples,” IEEE Trans.Image Process., vol. 15, no. 4, pp. 1017–1030, Apr. 2006.
Feature Weighting Query Moving
Existing Methods
• Subspace Learning for CBIR• X. Zhou and T. Huang, “Small sample learning during multimedia retrieval using biasmap,” in Proc. IEEE
Int. Conf. Computer Vision and Pattern Recognition (CVPR), 2001, vol. 1, pp. 11–17.
• D. Tao, X. Tang, X. Li, and Y. Rui, “Kernel direct biased discriminant analysis: A new content-based image retrieval relevance feedback algorithm,” IEEE Trans. Multimedia, vol. 8, pp. 716–727, 2005.
• D. Xu, S. Yan, D. Tao, S. Lin, and H.-J. Zhang, “Marginal fisher analysis and its variants for human gait recognition and content-based image retrieval,” IEEE Trans. Image Process., vol. 16, no. 11, pp.2811–2821, Nov. 2007.
Biased Discriminant Analysis
• Reasonable assumption:
– “All the positive samples are alike, but each negative sample is negative in each way.”