Transformed Principal Gradient Orientation for Robust and Precise Batch Face Alignment
This paper addresses the problem of simultaneously aligning a batch of linearly correlated images despite large misalignment, severe illumination and occlusion. Our algorithm assumes that the gradient orientation of images, if correctly aligned, can be robustly represented by an underlying transformed principal gradient orientation (TPGO) subspace. With such a linear representation prior, the proposed method connects PGO subspace learning, gradient orientation reconstruction, and batch alignment in a unified framework with an efficient alternating optimization solution. Besides inherent robustness from the gradient orientation and the low-rank structure, TPGO maintains the pixel-accurate registration precision and the efficient optimization of Lucas & Kanade framework. Experimental results show TPGO based batch alignment is more precise and robust than the state-of-the-art methods such as RASL and SIFT feature base Congealing. Moreover, integrated with a SIFT based pre-alignment procedure, TPGO is able to align a large number of images of multiple objects with large deviation, illumination, and occlusion in the precision that surpasses the handcrafted alignments (provided by the standard database distributions), in term of our face recognition experiments on the Extended Yale B, AR and FERET databases.
KeywordsGradient Orientation Sift Feature Subspace Learning FERET Database Automatic Face Recognition
This work was partially sponsored by National Natural Science Foundation of China (NSFC) under Grant No. 61375031, No. 61471048, and No. 61273217. This work was also supported by the Fundamental Research Funds for the Central Universities, Beijing Higher Education Young Elite Teacher Project, and the Program for New Century Excellent Talents in University.
- 8.Huang, G.B., Jain, V., Learned-Miller, E.: Unsupervised joint alignment of complex images. In: IEEE 11th International Conference on Computer Vision, 2007, ICCV 2007, pp. 1–8. IEEE (2007)Google Scholar
- 9.Cox, M., Sridharan, S., Lucey, S., Cohn, J.: Least squares congealing for unsupervised alignment of images. In: IEEE Conference on Computer Vision and Pattern Recognition, 2008, CVPR 2008, pp. 1–8. IEEE (2008)Google Scholar
- 10.Frey, B.J., Jojic, N.: Transformed component analysis: joint estimation of spatial transformations and image components. In: The Proceedings of the Seventh IEEE International Conference on Computer Vision, 1999, vol. 2, pp. 1190–1196. IEEE (1999)Google Scholar
- 11.Schweitzer, H.: Optimal eigenfeature selection by optimal image registration. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1999, vol. 1. IEEE (1999)Google Scholar
- 13.Vedaldi, A., Guidi, G., Soatto, S.: Joint data alignment up to (lossy) transformations. In: IEEE Conference on Computer Vision and Pattern Recognition, 2008, CVPR 2008, pp. 1–8. IEEE (2008)Google Scholar
- 14.Peng, Y., Ganesh, A., Wright, J., Xu, W., Ma, Y.: RASL: robust alignment by sparse and low-rank decomposition for linearly correlated images. In: 2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 763–770. IEEE (2010)Google Scholar
- 19.Tzimiropoulos, G., Zafeiriou, S., Pantic, M.: Robust and efficient parametric face alignment. In: 2011 IEEE International Conference on Computer Vision (ICCV), pp. 1847–1854. IEEE (2011)Google Scholar
- 20.Martinez, A.M., Benavente, R.: The AR face database. CVC Technical report #24, June 1998Google Scholar