Learning with Multiple Orientations and Scales for Face Recognition
A novel feature fusion algorithm using multiple orientations and scales for illumination-robust face recognition is proposed in this paper. For a given image, it will firstly be transformed by a group of Gabor filters, and the transformed coefficients be weighted by a vector, which can be determined by a given discriminant criteria and constrained quadratic programming method, then the weighted sum of these vectors is defined as the feature representation of the facial image. The new method provides a framework to regularize and calculate the complex similarity between different scales and orientations features, and it effectively avoids dimensionality curse problem emerged in some concatenation based feature fusion methods. Meanwhile, our method presents a reasonable approach for sensing discriminant orientations and scales according to the optimized weights. The framework can be extended into general multi-modal pattern recognition problems. Experiments using benchmark databases show that our new method obtains competitive performance and improves the recognition result.
KeywordsFace Recognition Illumination Orientation Scale Regularization
Unable to display preview. Download preview PDF.
- 4.Liu, C.J., Wechsler, H.: Gabor feature based classification using the enhanced Fisher linear discriminant model for face recognition. IEEE TIP 11(4), 467–476 (2002)Google Scholar
- 5.Shen, L.L., Bai, L.: Gabor wavelets and kernel direct discriminant analysis for face recognition. In: ICPR, pp. 284–287 (2004)Google Scholar
- 7.Xie, X., Lam, K.: Gabor-based kernel PCA with doubly nonlinear mapping for face recognition with a single face image. IEEE TIP 15(9), 2481–2492 (2006)Google Scholar
- 8.Lei, Z., Chu, R., He, R., Liao, S., Li, S.: Face recognition by discriminant analysis with gabor tensor representation. Advances in Biometrics, 87–95 (2007)Google Scholar
- 11.Dai, D.Q., Yan, H.: Wavelets and face recognition. In: Delac, K., Girgic, M. (eds.) Face Recognition, pp. 59–74. I-Tech Education and Publishing (2007)Google Scholar
- 13.Yan, F., Mikolajczyk, K., Barnard, M., Cai, H., et al.: ℓp norm multiple kernel Fisher discriminant analysis for object and image categorisation. In: IEEE CVPR, pp. 3626–3632 (2010)Google Scholar