Automatic generation of realistic 3-D human head models remains a challenging
task in computer vision and computer graphics for many years. Various
applications like virtual reality, computer games, video conferencing, and
filmmaking could benefit from photorealistic human face models. Although
there exist devices like laser range scanners and structured light
installations that can capture accurate 3-D shape of complex objects,
they are very expensive and not easy to use. Our goal is to make p
hotorealistic human head models very easily for a common PC user.
This means using affordable devices (like digital camera) for data
acquisition, maximizing possible level of automation along with controls
over the creation process, and robust algorithms which generate plausible
results even in case of imperfect input data acquired with simple equipment
by a non-professional user.
The tasks of the project include:
- Face detection and facial features extraction from frontal and profile view;
- Intellectual 3D model morphing;
- 3D model reconstruction from noisy depth data;
- Texture generation;
This project was started in 2001 in cooperation with Samsung Advanced Institute of Technology.
The project lasted for three years. As a result, a
face modeling system that produces a polygonal textured head model from
frontal and profile images was created. Since it is designed to be used
by a common user with cheap equipments, it is highly automated and robust
to the conflicting data and different picture-taking conditions.
Now this research continues in Multimedia Lab., Samsung Advanced Institute
of Technology.
Illustrations:
|
Automatic feature detection
|
|
|
Source images and resulting models
|
Model examples:
Several examples of reconstructed models are available.
Use any VRML viewer (for example
Cortona VRML Client)
to view the models.
The project team:
Principal researcher -
Vladimir Vezhnevets (vvp@graphicon.ru)
Researchers: Dmitriy Yurkovets, Stanislav Soldatov, Lisitsin Eugene, Degtiareva Anna,
Krasavin Denis, Shumaeva Tatiana
Collaborators at the SAIT side: In Kyu Park (Multimedia Lab, SAIT)
Publications:
In Kyu Park, Hui Zhang, Vladimir Vezhnevets, and Heui-Keun Choh
"Image-based Photorealistic 3-D Face Modeling".
To be presented in 6th IEEE International Conference on Automatic Face and Gesture Recognition (FG04), Seoul, Korea, May 2004.
Vezhnevets V., Soldatov S., Degtiareva A., and Park I.-K.
"Automatic extraction of frontal facial features".
Proc. Sixth Asian Conference on Computer Vision (ACCV04), vol. 2, pp. 1020-1025, Jeju, Korea, January 2004.
.pdf (1706kb)
.ps.gz (~3Mb)
Vezhnevets V., Degtiareva A.,
"Robust and Accurate Eye Contour Extraction".
Proc. Graphicon-2003, pp. 81-84, Moscow, Russia, September 2003.
.pdf (429kb)
.ps.gz(~800kb)
|