Partially Occluded Face Recognition and Components Estimation

Face recognition techniques are user friendly and useful to identify people without disturbance. Under constrained conditions most of the face recognition systems give better result results as compared to natural conditions. When images are associated with occlusions, varying illumination, pose variation, the result of face recognition systems decrease. Presently partial occlusion is a huge problem for face recognition, which comes due to scarf, sunglasses. This paper explains how to complete face, which is associated with occlusion or some components are missing. Main steps for the completion of face are: individual components of occluded face are detected by using Viola and Jones extension proposed by Lienhart et al. [1] [2], and faces from database are used to match with these components using AdaBoost ensemble learning. Euclidean distance can be used for this purpose also. Finally occluded face completed by using graph matching techniques