3D Modeling and Facial Regions based Unconstrained Face Recognition

Abstract

Unconstrained face image recognition is one of the most challenging research issues in the field of computer vision and intelligent machines (IM) due to different problems such as variation in pose, expression, illumination, and resolution. Although several face recognition algorithms have been suggested by researchers, recognition of unconstrained images presents low recognition rates. This study presents an approach of 3D modeling of 2D unconstrained face images, using intelligent machine tools for pose correction and to increase the number of training images. Principal Component Analysis (PCA) and Fisher Linear Discriminant Analysis (FLDA) have been employed to extract features from various face regions to develop weighted PCA and weighted FLDA. Furthermore, the matching scores level fusion of these two techniques is used to enhance recognition accuracy. The experimental results on two standard face image databases, Surveillance Cameras face (SCface) and Labeled Faces in the Wild (LFW), show the highest rank-1 identification accuracy of 39.23% and 42.22%, respectively, which is better than state-of-the-art approaches.