Ant Colony Optimized Local Binary Pattern Technique in Face Recognition

In the face of rising concern for public information security, Biometric indicator is an attractive solution, which inducts Computer Vision in general and Face Recognition in particular as an emerging solution. Face recognition systems differs in face representation and matching approaches. This work presents a Local Binary Pattern texture analysis for facial representation which is independent of expression, pose and illumination artefacts by extracting micro primitive represented as statistics of gray differential. This transformation gives Ant Colony Optimization algorithm a suitable representation to formulate the optimization problem of feature selection as a graph where pixels represents nodes. With correlation coefficient used as the heuristic desirability of pixel and pixel intensity representing pheromone level, optimal feature subsets in form of reduced pixels were obtained, which is salient enough to recognize a human face. The operational dataset captured under an uncontrolled environment, which contains high intra-class variation validates the effectiveness of the technique for practical application.