Face recognition with expression variation via robust NCC

We introduce a novel algorithm namely Robust Normalized Cross-correlation Coefficient (RNCC) for 2D frontal face recognition with expression variation. There are thirteen renowned ways to look at the Cross-correlation Coefficient. Our proposed method makes use of the technique named "Correlation as a Rescaled Variance of the Difference between Standardized Scores". It is based on estimating robust correlation via the rejection of outliers which original Normalized Cross-correlation Coefficient (NCC) is not capable of. We tested our technique on 6 renowned databases (AR, Cohn Kanade, Cohn Kan ade plus, Yale Faces, Bosphorus, and Jaffe) and have obtained exceptionally remarkable results. The rigorous testing has been performed using minimum training samples and low dimensionality (13 × 10). The recognition rates for all the tested databases are above 90% and outperform existing techniques.