3-D Head Model Acquisition (1996)
        Home Gavrila
        Chamfer System
        Facial Features
        3D Head Model
        3D Body Tracking

The acquisition of 3-D textured head models from images is important in applications such as computer animation, virtual reality and special effects. For example, many recent movies have come to rely on high precision scanned 3-D head models for their special effects; twisting, bending and morphing an actor´s appearance to the degree desired.

Traditional scanners have invariably required controlled studio environments and expensive equipment, costing several thousands of dollars.

For my Ph.D. Thesis, I developed a system which uses a single video-camera for the acquisition of an accurate 3-D human head model. The system is meant for a low-cost, mass application where the general public can capture their own 3-D appearance. All is needed is a PC and a (cheap) camera.

So how does this work?

In the initialization phase, the system requires the user to look straight in the camera. From this frontal view, characteristic facial features, such as eyes and lips, are extracted. These are used to initialize a generic 3-D head model.



In addition to characteristic facial features, person-specific features are localized. These combined set of features are tracked over time, as the user freely turns his/her head sideways to the camera. During this phase, the system estimates both head motion and head shape, fine-tuning the generic head model as more image data becomes available. Here is a short video clip which illustrates part of this process, i.e. how the head contours are assembled.




Above two figures show a point-rendering of the 3-D head model as computed after one head turn (frontal and sideway view)


Finally, here is a textured rendering of the 3-D head model after one head turn.

[Home Gavrila] [Resume] [Research] [People] [Publications] [Datasets] [Media Coverage] [Open Positions] [Search]

Copyright © 2001-2006 Gavrila. All rights reserved. Disclaimer.