Interest-Point based Face Recognition System
César Fernández, María Asunción · InTech eBooks · 2010
Automatic or supervised surveillance applications impose strict requirements in face recognition algorithms, in terms of false alarm rate and percentage of undetected criminals. We present a novel method, based on interest point detectors (namely, Harris-Laplace and DoG) and SIFT descriptors. Our measure of similarity between images is based on computing the number of corresponding points that, apart from having similar values for their SIFT descriptors, fulfil scale, orientation and relative location coherence. Images with a higher number of