Illumination Invariant Face Recognition

In this book an in-depth study of the state-of-art illumination invariant face recognition techniques have been carried out and a method based on the fusion of two different feature extraction techniques is proposed to overcome the adverse illumination conditions. The proposed system uses the gradient based illumination normalization to remove the illuminance component superiority. To obtain the illumination insensitive face representation, a ratio of the gradient amplitude to the original image intensity is obtained. The facial features are extracted using two different feature extraction techniques. Local binary pattern (LBP) is a very efficient local texture descriptor based on thresholding the pixels in a small neighborhood based on the value of the center pixel. Local ternary pattern (LTP) is a noise resistant modified version of LBP. The features vectors provided by the two techniques are fused at feature level. Finally artificial neural network is used in the classification stage for recognition purpose.