Local Centre of Mass Face for face recognition under varying illumination

In this work we propose a novel method to extract illumination insensitive features for face recognition called local centre of mass face (LCMF). In this LCMF approach the gradient angle between the centre of mass and centre pixel of a selected neighborhood is extracted. Theoretically it is shown that this feature is illumination invariant using the Illumination Reflectance Model (IRM) and is robust to different illumination variations. It is also shown that this method does not involve any explicit computation of Luminance (L) component and as centre of mass is an inherent feature of a mass distribution, its slope with the centre pixel of the neighborhood has local edge preserving capabilities. The angle of the slope obtained using Centre of Mass with the centre pixel of the neighborhood is used as a feature vector. This feature vector is directed from the darkest section of the neighborhood to the brightest section of the neighborhood as Centre of Mass is always positioned towards the brighter side of a mass distribution and hence encrypts the edge orientation. Using the L-1 norm distance measure, these feature vectors are used to classify the images. The method does not involve any preprocessing and training of images. The proposed method has been successfully tested under different illumination variant databases like AR, CMU-PIE, and extended Yale B using standard protocols, and performance is compared with recently published methods in terms of rank-1 recognition accuracy. The method is also applied on Sketch-Photo pair database like CUHK. For unbiased or fair performance evaluation, the Sensitivity and Specificity are also being measured for the proposed method on all the databases. The proposed method gives better accuracy performance and outperforms other recent face recognition methods.