Laser scan quality 3-D face modeling using a low-cost depth camera

We propose a method to produce laser scan quality 3-D face models from a freely moving user with a low-cost, low resolution depth camera. Our approach does not rely on any prior face model and can produce faithful geometric models of star-shaped objects. We represent the object in cylindrical coordinates, which enables us to perform filtering operations very efficiently. We initialize the model with the first depth image, and then register each subsequent cloud of 3-D points to the reference using a GPU (Graphics Processing Unit) implementation of the ICP (Iterative Closest Point) algorithm. This registration is robust in that it rejects poor alignment due to facial expressions, occlusions, or a poor estimation of the transformation. We perform both temporal and spatial smoothing of the successively incremented model. To validate our approach, we quantitatively compare our model to one produced by laser scanning, and show comparable accuracy.