Minimizing Separability: A Comparative Analysis of Illumination Compensation Techniques in Face Recognition

Feature extraction task are primarily about making sense of the discriminative features/patterns of facial information and extracting them. However, most real world face images are almost always intertwined with imaging modality problems of which illumination is a strong factor. The compensation of the illumination factor using various illumination compensation techniques has been of interest in literatures with few emphasis on the adverse effect of the techniques to the task of extracting the actual discriminative features of a sample image for recognition. In this paper, comparative analyses of illumination compensation techniques for extraction of meaningful features for recognition using a single feature extraction method is presented. More also, enhancing red, green, blue gamma encoding (rgbGE) in the log domain so as to address the separability problem within a person class that most techniques incur is proposed. From experiments using plastic surgery sample faces, it is evident that the effect illumination compensation techniques have on face images after pre-processing is highly significant to recognition accuracy.