An improved ABC algorithm approach using SURF for face identification

Abstract

Face recognition is being intensively studied in the areas of computer vision and pattern recognition. Working on still images with multiple faces is a challenging task due to the inherent characteristics of the images, the presence of blur, noise and occlusion, as well as variations of illumination, pose, rotation and scale. Besides being invariant to these factors, face recognition systems must be computationally efficient and robust. Swarm intelligence algorithms can be used for object recognition tasks. Based on this context, we propose a new approach using an improved ABC implementation and the interest point detector and descriptor SURF. To assess the robustness of our approach, we carry out experiments on images of several classes subject to different acquisition conditions. © 2012 Springer-Verlag.