Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Learning group-based dictionaries for discriminative image representation
Lei H., Mei K., Zheng N., Dong P., Zhou N., Fan J. Pattern Recognition47 (2):899-913,2014.Type:Article
Date Reviewed: May 27 2014

Being able to correctly find and identify images based on either visual or language cues is an activity easily performed by humans. Unsupervised computing environments have yet to reach the same level of accuracy, as is evident from the image search engines available. Nevertheless, over the last decade, the performance of these search engines has improved markedly.

The reason for these advances is the continuous development and improvement of algorithms focused on automated image feature extraction and classification. In 2003, Sivic and Zisserman proposed a new image feature classification borrowed from text-based classification that considers images as sets of feature values [1]. Add extremely efficient algorithms that define and extract image features (points of interest), also developed recently, and image classification is within reach. The problem is complex, though, given the many different visual classes images are grouped in, the overlap in features and in meanings, and the high computational costs. In this paper, Lei et al. provide improvements on both feature representation and image classification by considering groups of similar visual classes. The similarity among the image classes is not determined through their labels, but is automatically computed using distances on the same features that will ultimately lead to image classification. Adding this grouping layer allows for a better differentiation that, in turn, improves classification compared to previously introduced techniques.

The algorithms are explained in a clear fashion that could also be improved by the availability of sample code. The research is supported by convincing experiments that include well-established image datasets and comparisons with known prior techniques. Especially important in situations where large datasets are employed is a good understanding of the computational costs of the new algorithm. The authors provide execution times for various runs and data sizes, yet they do not expand the empirical observations into a theoretical characterization of computational complexity.

Reviewer:  Stefan Robila Review #: CR142314 (1408-0690)
1) 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 IEEE, 2003, 1470–1477.
Bookmark and Share
  Featured Reviewer  
 
Image Representation (I.4.10 )
 
 
Dictionaries (H.3.1 ... )
 
 
Learning (I.2.6 )
 
Would you recommend this review?
yes
no
Other reviews under "Image Representation": Date
On detecting all saddle points in 2D images
Kuijper A. Pattern Recognition Letters 25(15): 1665-1672, 2004. Type: Article
Jul 14 2005
General adaptive neighborhood image processing
Debayle J., Pinoli J. Journal of Mathematical Imaging and Vision 25(2): 267-284, 2006. Type: Article
Mar 29 2007
Human skeleton tracking from depth data using geodesic distances and optical flow
Schwarz L., Mkhitaryan A., Mateus D., Navab N. Image and Vision Computing 30(3): 217-226, 2012. Type: Article
Aug 26 2013
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy