Inexpensive construction of a 3D face model from stereo images

Abstract

Construction of a three dimensional face model from stereo images is a challenging task. Most of the currently available systems for reconstruction of 3D models require special hardware for calibration. In this paper, we illustrate a mechanism to construct a three dimensional face model from two stereo images. The developed mechanism does not require any special devices to calibrate the stereo images. We used a hand-held inexpensive digital camera to take the stereo images of a face. We did not use any camera-stand to fix and measure the camera system geometry. The stereo images were taken holding the camera in hand and moving it to two slightly different viewpoints. We constructed a depth map from these two stereo images and utilized this depth map to reconstruct the three dimensional face model. The 3D face model reconstruction process described in this paper uses some existing theories and combines them to develop a new system to generate the depth map. The system requires minimal user interaction for the reconstruction.