PROJECT TITLE :

Face Recognition With Pose Variations and Misalignment via Orthogonal Procrustes Regression

ABSTRACT:

A linear regression-based method is a hot topic in face recognition community. Recently, sparse representation and collaborative representation-based mostly classifiers for face recognition have been proposed and attracted nice attention. However, most of the prevailing regression analysis-based mostly ways are sensitive to pose variations. During this paper, we tend to introduce the orthogonal Procrustes drawback (OPP) as a model to handle create variations existed in 2D face images. OPP seeks an optimal linear transformation between 2 pictures with completely different poses thus as to create the reworked image best fits the opposite one. We have a tendency to integrate OPP into the regression model and propose the orthogonal Procrustes regression (OPR) model. To handle the problem that the linear transformation isn't appropriate for handling highly non-linear pose variation, we have a tendency to additional adopt a progressive strategy and propose the stacked OPR. As a practical framework, OPR will handle face alignment, pose correction, and face representation simultaneously. We have a tendency to optimize the proposed model via an economical alternating iterative algorithm, and experimental results on three common face databases, such as CMU PIE database, CMU Multi-PIE database, and LFW database, demonstrate the effectiveness of our proposed technique.