Abstract
When we seek to directly learn basis functions from natural scenes, we are confronted with the problem of simultaneous estimation of these basis functions and the coefficients of each image (when projected onto that basis). In this work, we are mainly interested in learning matrix space basis functions and the projection coefficients from a set of natural images. We cast this problem in a joint optimization framework. The Frobenius norm is used to express the distance between a natural image and its matrix space reconstruction. An alternating algorithm is derived to simultaneously solve for the basis vectors and the projection coefficients. Since our fundamental goal is classification and indexing, we develop a matrix space distance measure between images in the training set. Results are shown on face images and natural scenes.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
C. Bishop. Neural Networks for Pattern Recognition. Oxford University Press, Oxford, 1995.
H. Chui and A. Rangarajan. A new algorithm for non-rigid point matching. In IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), volume 2,pages 44–51, 2000.
I. Daubechies. Ten lectures on wavelets. SIAM, 1992.
J. Daugman. Complete discrete 2D Gabor transforms by neural networks for image analysis and compression. IEEE Trans. Acoustics, SPeech and Signal Proc.7:1169–1179, July 1988.
G. Golub and C. Van Loan. Matrix Computations. Johns Hopkins University Press, 2nd edition, 1989.
M. Lewicki and B. Olshausen. Probabilistic framework for the adaptation and comparison of image codes. J. Opt. Soc. America,. 16(7):1587–1601, 1999
B. Olshausen and D. Field. Natural image statistics and efficient coding. Network, 7:333–339, 1996.
B. Olshausen and K. Millmann. Learning sparse codes with a mixture of Gaussians prior. In Advances in Neural Information Processing Systems (NIPS) 12, pages 841–847. MIT Press, Cambridge, MA, 2000.
K. Rao and P. Yip. Discrete Cosine Transform—Algorithms, Advantages and Applications. Academic Press, London, UK, 1990.
J. Yang and C. Lu. Combined techniques of singular value decomposition and vector quantization for image coding. IEEE Trans. Image Proc., 4:1141–1146, 1995.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2001 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Rangarajan, A. (2001). Learning Matrix Space Image Representations. In: Figueiredo, M., Zerubia, J., Jain, A.K. (eds) Energy Minimization Methods in Computer Vision and Pattern Recognition. EMMCVPR 2001. Lecture Notes in Computer Science, vol 2134. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-44745-8_11
Download citation
DOI: https://doi.org/10.1007/3-540-44745-8_11
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-42523-6
Online ISBN: 978-3-540-44745-0
eBook Packages: Springer Book Archive