3DLBP and HAOG fusion for face recognition utilizing kinect as a 3D scanner

Abstract

Pose and illumination variability are two major problems with 2D face recognition. Since 3D data is less sensible to illumination changes and can be used to adjust pose variations, it has been adopted to improve performance on face recognition systems. The main problem with utilizing 3D data is the high cost of the traditional 3D scanners. The Kinect is a low cost device that can be used to obtain the 3D data from an environment in a fast manner, but with lower accuracy than the traditional scanners. Recently, a 3D Local Binary Pattern (3DLBP) method was proposed for 3D face recognition by using high resolution scanners. The main goal of this work is to assess the performance of 3DLBP method, fused with Histogram of Averaged Oriented Gradients (HAOG) face descriptor method, for face recognition when Kinect is used as the 3D face scanner. Another goal is to compare the 3DLBP method, fused with HAOG descriptor, with other methods proposed in the literature for face recognition by using Kinect. Experimental results on EURECOM face dataset showed that the data generated by Kinect are discriminative enough to allow face recognition and that 3DLBP performs better than the other methods. Copyright is held by the owner/author(s).