Pose-Invariant Face Recognition in Video Sequences by 3D Face Reconstruction and Synthesis Face Pose

In this paper, a novel manner for unrestrained pose-invariant face recognition was proposed. Also, a novel and efficient method was proposed to reconstruct the 3D models of a human face from a single 2D face image with variety in facial expression using the Deformable Generic Elastic Model (D-GEM). Three generic models were em-ployed for modeling facial expression in the Generic Elastic Model (GEM) framework and a mixture of these three models by using computing distance around face lips. Par-ticularly, present method was tested on an available 2D face databases without facial expression images and a new synthesized sequences pose from gallery images and com-pared present synthesized results with target face images by performing face recognition using the rank-one recognition rate with the smallest cosine distance. Promising results were acquired for handling pose changes based on the proposed method compared to the GEM approach.