Face Recognition in Uncontrolled Environment

This paper presents a novel method of facial image representation for face recognition in uncontrolled environment. It is named as augmented local binary patterns (A-LBP) that works on both, uniform and non-uniform patterns. It replaces the central non-uniform pattern with a majority value of the neighbouring uniform patterns obtained after processing all neighbouring non-uniform patterns. These patterns are finally combined with the neighbouring uniform patterns, in order to extract discriminatory information from the local descriptors. The experimental results indicate the vitality of the proposed method on particular face datasets, where the images are prone to extreme variations of illumination. Keywords—Face recognition, A-LBP, descriptor, distance metrics, area under curve, decidability index