Face localization using skin colour and maximal entropy based particle swarm optimization for facial recognition

One amongst the vital pre-processing operation in every Face Recognition System is Face localization. There is a growing need for algorithms that quickly yield the correct region of interest around the face. Here an algorithm is proposed which performs skin detection using Particle Swarm Optimization for face localization. Where minimization of fitness function is used to resolve optimization problems. Here bounding boxes are modelled as particles, or randomly distributed test vectors to ascertain for the region of interest in an image where it is likely that the face exists, with the fitness value returned by the fitness function that tests the interior of a box for maximum entropy and maximum skin like features, which in combination indicates how likely the interior represents a face. Consequently, the particle (bounding box) with the best fitness value is chosen as face region of the image, and the output is fed to a Binary Particle Swarm Optimization based Face Recognition System therefore yielding higher recognition rates than conventional systems.