Use of matrix polar decomposition for illumination-tolerant face recognition in discrete cosine transform domain

• DOI: 10.1049/iet-ipr.2009.0340
• Publication Date: 2011-02-01T02:10:09Z
• Authors: M. Ezoji; K. Faez
• Keywords: Matrix (chemical analysis)

In this study, an illumination-tolerant face recognition algorithm is proposed. This work highlights the significance of matrix polar decomposition for illumination-invariant recognition. The proposed has two stages. first stage, authors reduce effect illumination changes by weakening discrete cosine transform coefficients block intensities using a new designed quantisation table. second unitary factor reconstructed image used as feature matrix. phase, novel indirect method measuring similarities in matrices nearest-neighbour rule applied to matching. have performed some experiments on several databases evaluate its different aspects. Experimental results demonstrate that approach provides suitable representation invariant