Comparison of Algorithms for the Detection of Occluded Faces

Abstract

Due to growing demands in various application areas face recognition has attracted greater attention in the recent years. The advantages of using face as an identity over any other alternate methods, such as fingerprint identification, is relying on the fact that it is quite easy to use and is well accepted by the people as it doesn’t require any physical interaction of the user and the system and generates better result. Facial recognition is not dynamic enough to be used in practical security applications because of some problems associated with it and these are variations in pose and lightning conditions. During last few years, different algorithms have been designed to overcome these problems but very few have been designed in order to address the problem of partial occlusions. Occlusion detection in face verification is an essential problem nowadays. Occlusion in face recognition is common in today’s world but yet a challenging problem. We tried to build a system that could deal with images with occlusion and the another problem which was related to varying pose was tried to be minimised and the implemented algorithm has shown better results with varying pose. The challenging problem where our system fails is related to variation in illumination that is system gives false results when images are tested that were clicked in poor lightning conditions but still system is somewhat robust enough when it comes to security applications.