Design of a Face Recognition Model using Periocular Region for Identification of Masked Faces

Abstract

The most emerging field of computer vision is facial recognition where the newlineidentification of individual has captured the interest of many scientists. For more than four newlinedecades the area of face recognition has emerged as a widespread metric for assessing human newlineidentification. Reliability of the existing facial recognition systems still has a lot of challenges newlinewith respect to computer vision and pattern recognition irrespective of the various newlineadvancements made. The challenging part in recognizing faces occur mostly in scenarios like newlinerecognizing twins face, different profile views of face, having different accessories like glasses, newlinebeard, etc. The surrounding conditions also plays a major role like light intensity, illumination newlinecondition, occlusion, etc. However, the performance of face recognition systems declines if the newlineface is partially hidden. In different situations, face is either covered by helmets, hair, glasses newlineor skiing masks. In most of these cases, the region around the eyes (periocular) is the only newlinevisible trait which can be used as biometric. newline newlineThe main objective of this work is to develop a novel efficient framework for identifying human newlineusing periocular biometric. Several periocular biometric detection approaches are already newlinedeveloped in earlier. But all of the well-established approaches are not robust enough to handle newlinevariations in resolution, illuminated images, postures due to the inefficiency of hand-crafted newlinetexture descriptors, especially while working with periocular features. Thus, periocular newlinerecognition still paves a way for a lot of research enhancements. newline