Two-Dimensional Partial Least Squares and Its Application in Image Recognition
The problem of extracting optimal discriminant features is a critical step in image recognition. The algorithms such as classical iterative partial least squares (NIPALS and CPLS), non-iterative partial least squares based on orthogonal constraints (NIPLS), and partial least squares based on conjugation orthogonal constraints (COPLS) are introduced briefly. NIPLS and COPLS methods based on original image matrices are discussed where image covariance matrix is constructed directly using the original image matrices just like 2DPCA and 2DCCA. We call them 2DNIPLS and 2DCOPLS in the paper. Two arbitrary optimal discriminant features can be extracted by 2DCOPLS based on uncorrelated score constraints in theory. At the same time, it is pointed out that 2DCOPLS algorithm is more complicated than other PLS based algorithms. The results of experiments on ORL face database, Yale face database, and partial FERET face sub-database show that the 2DPLS algorithms presented are efficient and robust.
KeywordsPartial Least Squares (PLS) Uncorrelated Constraints 2DPCA Optimal Projection Image Recognition
Unable to display preview. Download preview PDF.
- 1.Wold, H.: Estimation of Principal Components and Related Models by Iterative Least Squares. In: Multivariate Analysis. Academic, New York (1966)Google Scholar
- 4.Wold, H.: Path with Latent Variables: The NIPALS Approach. In: Balock, H.M. (ed.) Quantitative Sociology: International Perspectives on Mathematical and Statistical Model Building, pp. 307–357. Academic Press, London (1975)Google Scholar
- 7.Yang, J., Yang, J.-Y., Jin, Z.: A Feature Extraction Approach Using Optimal Discriminant Transform and Image Recognition. Journal of Computer Research & Development 38, 1331–1336 (2001)Google Scholar
- 9.Han, L.: Kernel Partial Least Squares for Scientific Data Mining. PHD thesis, Rensselaer Polytechnic Institute, Troy, New York (2007)Google Scholar
- 10.Arenas-García, J., Petersen, K.B., Hansen, L.K.: Sparse Kernel Orthonormalized PLS for Feature Extraction in Large Data Sets. In: Advances in Neural Information Processing Systems, vol. 19. MIT Press, Cambridge (2007)Google Scholar
- 12.Jacob, A.: A Survey of Partial Least Squares Methods, with Emphasis on The Two-block Case. Technical Report, Department of Statistics, University of Washington, Seattle (2000)Google Scholar
- 16.Bolme, D.S., Beveridge, J. R., Teixeira, M., Draper, B. A.: The CSU Face Identification Evaluation System: Its Purpose, Features, and Structure. In: Proceedings of 3rd International Conference on Computer Vision Systems (ICVS), pp.304–313 (2003)Google Scholar