Facial Recognition has gained interested of many researchers due to its various purposes and practical applications.
Recent developments in 3D imaging technology have made it possible to use a true 3D mesh of the face for recognition.
Moreover, it is now also possible to use 4D data (3D meshes changing over time) from publicly-available databases.
Since the use of 3D and 4D data promises to bypass many of the issues that plague 2D images and video, such as pose and
facial expression variations, the challenge is to perform expression analysis on 3D and 4D domains to achieve high face
recognition rate. The key contribution of this article is two fold, firstly, we present various 3D Face Recognition techniques
and secondly, we present a systematic pipeline for 4D Face Recognition (dynamic 3D sequences) which consist of
Mesh Matching, RANdomSAmple Consensus (RANSAC), consecutive Annotated Face Model (AFM) Fitting and Feature
Description. We evaluated both latest state of the art 3D and 4D Facial Recognition pipelines on two publicly available
facial expression databases BU-3DFE and BU-4DFE and made the following observations. Among various 3D face recognition
algorithms analysed Matching Tensors for Pose Invariant Automatic 3D Face Recognition gives the best recognition
rate of 96.4% on BU-3DFE. Among various 4D face recognition algorithms, a dynamic geometry-based approach for
4D facial expressions recognition gives the best recognition rate of 96.71% on BU-4DFE. We have also considered several
patents, such as ‘Face recognition apparatus, face recognition method, Gabor filter application apparatus, and computer
program’ ‘Gabor filtering and joint sparsity model-based face recognition method’ ‘Principal component analysisbased
(PCA-based) three-dimensional (3D) face recognition system’ ‘3d face recognition method based on intermediate
frequency information in geometric image’ ‘Three-dimensional human face recognition method based on intermediate
frequency information in geometry image’.
Annotated face model, 3D mesh, 3D morphable models, face recognition, feature description, mesh matching, random
Research Scholar, Electronics & Communication Engineering, Karunya University, Coimbatore, India.