3D face recognition using stereo images

Abstract

Face recognition in still images and video is an important application in computer vision and image processing. This work focuses on extracting 3D or depth information using stereo matching for the purpose of enhancing face recognition. We present a new fast stereo matching algorithm termed stereo cluster search (SCS) that can be performed in real time to generate depth information from the left and right stereo pair of a face image. The depth information is then augmented with the 2D face image information for enhancing recognition using Principal Component Analysis. The work uses a 3D labeled database of face images for testing the proposed method that proves suitable for the recognition problem.