Illumination Invariant Face Recognition of Newborn Using Single Gallery Image

Face recognition of newborn babies under different illumination condition is one of the key challenges. There have been many methods proposed by the researchers to handle this face recognition problem, but most of them are using more than one gallery image for building up the model. But in the case of newborn, capturing the image in similar illumination condition is sometimes not feasible, moreover their facial expressions and the pose are naturally uncontrolled. We have proposed a model to handle the problem of illumination for newborn using a single gallery image. The proposed method uses a combined approach of local binary pattern (LBP) for feature extraction and semi-supervised learning. The results show that identification accuracy in case of newborn images is improved by 11 % compared to simple LBP with 81 % for rank 1.