Non-Frontal Low Resolution Face Image Recognition Based on Canonical Correlated Principal Component Analysis

Non-Frontal Low Resolution Face Image Recognition Based on Canonical Correlated Principal Component Analysis
Authors:KARTHIK.T, RAJENDRAPRASAD.D

Abstract: In video surveillance, the face recognition usually aims at recognizing a non-frontal low resolution face image from the gallery in which each person has only one high resolution frontal face image. Traditional face recognition approaches have several challenges, such as the difference of image resolution, pose variation and only one gallery image per person. This paper proposes a new method for face recognition in the case of “one sample per class” using one non-frontal LR input. FH features are super resolved from NFL input by the learnt nonlinear mappings in the coherent space. The nonlinear regression models from the specific non-frontal low resolution image to frontal high resolution features are learnt by radial basis function in subspace built by canonical correlation analysis. Extensive experiments on benchmark database show the superiority of our method. 

Keywords: Canonical Correlation Analysis, Non-Frontal Face Recognition, Radial Basis Function, Super Resolution.

 INTRODUCTION 

        Although human beings can easily detect and identify faces in a scene, it is very challenging for an automated system to achieve such objectives. Face recognition has drawn great attention in recent decades, due to its wide range commercial and law-enforcement applications [1]. The challenges become more profound when large variations exist in the face images at hand, e.g., variations in illumination conditions, viewing directions or poses, facial expression, aging, and disguises such as facial hair, glasses, cosmetics and scarves. Despite of these challenges, face recognition has drawn wide attention from researchers in areas of machine learning, computer vision, pattern recognition, neural networks, and so on. Super resolution methods are used for LR Face recognition [2], [3], [4], [5], [6]. In this paper, this work mainly focus on improving the recognition performance in the case where only a single face “snapshot” of LR is available. 

                                                                                  Read More....