Hybrid methodology focused on the model of binary patterns and the theory of fuzzy logic for facial biometric verification and identification

It is proposed a methodology to improve verification and identification biometric facial indicators based in hybridization binary pattern models and the fuzzy logic theory, making besides use of the traditional image pre-processing models, feature extraction and classifiers to validate the performance of the proposal methodology. The facial recognition is complicated due to the variability of the facial appearance related the same person, and the small characteristic samples for each person in adverse conditions. To fix this, is considered the binary pattern models as an excellent choice to the local face representations, whose more important properties is their tolerance against the variations of luminance, scale and rotation. However, the binary pattern model is sensitive to small variations of the pixel intensities, generally caused by the noise, which introduce uncertainty to the texture and contrast representation of the facial image. Using fuzzy logic in the binary patterns calculation, leads to a texture representation model that takes into account the uncertainty of the contained information in each image, providing a better representation of texture and contrast measure. In combination with traditional algorithms in the pre-processing stage, as photometric and histogram normalization, the feature extraction stage is achieved using linear discriminants and Gabor wavelets to provide finally a stage of the support vector machines classification.