Three-Dimensional Face Recognition in the Presence of Facial Expressions: An Annotated Deformable Model Approach

Abstract

Abstract-In this paper, we present the computational tools and a hardware prototype for 3D face recognition. Full automation is provided through the use of advanced multistage alignment algorithms, resilience to facial expressions by employing a deformable model framework, and invariance to 3D capture devices through suitable preprocessing steps. In addition, scalability in both time and space is achieved by converting 3D facial scans into compact metadata. We present our results on the largest known, and now publicly available, Face Recognition Grand Challenge 3D facial database consisting of several thousand scans. To the best of our knowledge, this is the highest performance reported on the FRGC v2 database for the 3D modality.Index Terms-Face and gesture recognition, information search and retrieval.

Expression-insensitive 3D face recognition using sparse representation

“…First, the absence of a consistent parameterization among face meshes or point clouds makes it impossible to directly obtain registered features with a uniform sampling pattern; this prevents appearance-based methods or the sparse representation framework from being applied. More importantly, facial expressions can cause severe global geometry variations, resulting in degraded recognition performance [3,17] or heavy computational load [4,12].…”

Using 3D Models for Improving Face Recognition

“…Besides, this cannot be applied to 2D images, so it is not possible to work with existent databases. Some papers using this approach are [3] and [5].…”

Robust binary neural networks based 3D Face detection and accurate face registration

“…Faltemier et al 10 proposed a method for curvature and shape index based nose tip detection to localize the position of the nose tip and then align the whole input image to a template using the ICP algorithm. Kakadiaris et al 16 implemented a multistage alignment method including three algorithmic steps: Spin-images based alignment, ICP-based alignment and Simulated Annealing on Z-Buffers alignment. However, both of these approaches used the whole face area during their alignments.…”