Kernel-based distance metric learning for content-based image retrieval
Hong Chang,Dit-Yan Yeung +1 more
TLDR
This paper proposes a kernel approach to improve the retrieval performance of CBIR systems by learning a distance metric based on pairwise constraints between images as supervisory information and defines the transformation in the kernel-induced feature space which is nonlinearly related to the image space.About:
This article is published in Image and Vision Computing.The article was published on 2007-05-01 and is currently open access. It has received 55 citations till now. The article focuses on the topics: Metric (mathematics) & Content-based image retrieval.read more
Figures
Figure 4: Retrieval results on the first image database based on a separate set of query images. (a) cumulative neighbor purity curves; (b) precision/recall curves. 
Figure 5: Functional diagram for metric learning in CBIR. 
Figure 1: Retrieval results on the first image database (1010 images, 10 classes). (a) cumulative neighbor purity curves; (b) precision/recall curves. 
Figure 2: Retrieval results on the second image database (547 images, 6 classes). (a) cumulative neighbor purity curves; (b) precision/recall curves. 
Figure 7: Retrieval results based on stepwise kernel-based metric learning with different maximum batch sizes. (a) ω = 10 sessions; (b) ω = 20 sessions; (c) ω = 40 sessions. 
Figure 3: Typical retrieval results on the two databases ((a) and (b)) based on Euclidean distance (top row), DistBoost (middle row) and our kernel method (bottom row). Each row shows the 7 nearest neighbors including the query image (framed).
Citations
More filters
Proceedings ArticleDOI
Deep Learning for Content-Based Image Retrieval: A Comprehensive Study
TL;DR: This paper investigates a framework of deep learning with application to CBIR tasks with an extensive set of empirical studies by examining a state-of-the-art deep learning method (Convolutional Neural Networks) for CBIr tasks under varied settings.
Journal ArticleDOI
Locality constraint distance metric learning for traffic congestion detection
TL;DR: A Locality Constraint Distance Metric Learning (LCML) which ensured the local smoothness and preserved the correlations between samples is proposed for traffic congestion detection and extensive experiments confirm the effectiveness of the proposed method.
Proceedings ArticleDOI
Conference on Computer Vision and Pattern Recognition
TL;DR: Welcome to San Diego, California and the 2005 IEEE Society Conference on Computer Vision and Pattern Recognition (CVPR), held on June 20-26, where all of the proceedings are presented on a single DVD.
Journal ArticleDOI
Fast neighborhood component analysis
TL;DR: Experimental results show that, compared with NCA, FNCA not only significantly increases the training speed but also obtains higher classification accuracy, and comparative studies with the state-of-the-art approaches on various real-world datasets also verify the effectiveness of the proposed linear and nonlinear F NCA methods.
Journal ArticleDOI
A Kernel Approach for Semisupervised Metric Learning
Dit-Yan Yeung,Hong Chang +1 more
TL;DR: Experimental results show that this new kernel approach for semisupervised metric learning is promising for nonlinear metric learning.
References
More filters
Journal ArticleDOI
Nonlinear component analysis as a kernel eigenvalue problem
TL;DR: A new method for performing a nonlinear form of principal component analysis by the use of integral operator kernel functions is proposed and experimental results on polynomial feature extraction for pattern recognition are presented.
Journal ArticleDOI
Content-based image retrieval at the end of the early years
TL;DR: The working conditions of content-based retrieval: patterns of use, types of pictures, the role of semantics, and the sensory gap are discussed, as well as aspects of system engineering: databases, system architecture, and evaluation.
Journal ArticleDOI
Color indexing
Michael J. Swain,Dana H. Ballard +1 more
TL;DR: In this paper, color histograms of multicolored objects provide a robust, efficient cue for indexing into a large database of models, and they can differentiate among a large number of objects.
Proceedings Article
Distance Metric Learning with Application to Clustering with Side-Information
TL;DR: This paper presents an algorithm that, given examples of similar (and, if desired, dissimilar) pairs of points in �”n, learns a distance metric over ℝn that respects these relationships.
Proceedings Article
Constrained K-means Clustering with Background Knowledge
TL;DR: This paper demonstrates how the popular k-means clustering algorithm can be protably modied to make use of information about the problem domain that is available in addition to the data instances themselves.