ABSTRACT
ABSTRACT: The biggest challenge faced by Face Recognition is to extract important facial features efficiently. Existing methods like SIFT, Root SIFT perform quite accurately, but fail in real world applications that require real-time processing. The ORB feature detector-descriptor performs quite well in this regard. We have modified the ORB feature detector to Root ORB and used it as a pre-processing step for a study of classification algorithms for face recognition. Our results show that the Root SIFT feature detector-descriptor performs only 6.25% to 12.5% as fast as the Root ORB detector-descriptor. We have chosen to work with detector-descriptors like ORB rather than using Neural Networks to extract facial features. Compute-intensive methods like Neural Networks require a lot of training data and computational time to obtain these features.
