Illumination invariant face recognition system
C. Tosik, Alaa Eleyan, Mohammad Salman · 2013
One of the main issues that degrades the performance of a face recognition system is the illumination problem. In this paper, we investigated different algorithms as a preprocessing step to overcome the illumination problem. Histogram equalization, discrete cosine transforms and steerable Gaussian filters were applied to face images from Yale face database B for illumination normalization. PCA algorithm was used as a feature extractor. Using this preprocessing step the performance of the system shows significant improvements.