Hybrid Descriptor and Patches Optimization for Face Recognition

This work focuses on the system optimization for face recognition (FR) in controlled and uncontrolled environments. The strategy adopted in this article represents the contribution of hybrid and face patches descriptors for the extraction of facial features. The extracted features are optimize by the selection of optimal characteristics. In order to significantly improve the performance of system based on a single descriptor using the combination of several pieces of information and their optimization the Particle Swarm Optimization (PSO) algorithm method is used for optimize information. In this work, the problem of multiclass patches recognition is converted into face recognition. For each patch, a new feature optimization model based patch optimization is proposed to represent the specificity of the face expression. The patch representation-based approach is then proposed to detect the active region of interest of the testing sample for better generalization. The algorithm achieved competitive accuracies of real-world data: Labeled Faces in the Wild (LFW), as well as on Databases "Extended Yale face database B" and give a good performance.