Rotation distortions for improvement in face recognition with PCNC

Abstract

The face recognition is a very important task in security (airports, institutions, and so on) and authentication through photo tagging in social networks. We propose to improve face recognition with the Permutation Coding Neural Classifier (PCNC) using a special type of distortions of original images (for example, rotations) to train the neural network. We applied the distortions to the initial image database (the FRAV2D image database) and produced an extended rotated version of it that allowed us to improve the training process of PCNC neural classifier. The results obtained show a better recognition rate in comparison to Support Vector Machine (SVM) and Iterative Closest Point (ICP).