Two-year post-doctoral research position in multiple-camera reconstruction

Contact person: Radu Patrice HORAUD <Radu.Horaud<στο>inria.fr>

The PERCEPTION group has recently started a multi-year collaboration with the Samsung Advanced Institute of Technology, Seoul, South Korea.

Within this research project, PERCEPTION develops methods and algorithms for the visual reconstruction of a complex scene using a

mixed camera system. The camera system is composed of four units; each unit comprises one time-of-flight (TOF) low-resolution camera and two

high- resolution color cameras, a TOF-stereo sensor. The system is able to acquire finely synchronized multiple videos at 30 frames/second. We

already developed a method for calibrating each TOF-stereo unit [1] as well as a method  for calibrating the whole camera system.

We also developed a reconstruction method that starts with the 3D data provided by the four TOF cameras, followed by mesh refinement and

rendering using the eight color cameras and based on the mesh-evolution method described in [2].

The post-doctoral researcher will further continue to explore the potential of such a mixed camera system. In particular he/she will

investigate how to combine classical stereoscopic vision algorithms with the TOF sensors in order to obtain a high resolution dense depth map -

depth data upsampling. He/she will also investigate how to fuse TOF-data with stereo-data in order to deal with situations where depth

cameras alone or color cameras alone are unable to provide accurate/dense depth maps, e.g., scattering surfaces, non-Lambertian surfaces,

uniformly colored surfaces (no texture), bad lighting conditions, etc.

The expected outcomes of this research project are: scientific publications, patent applications jointly with Samsung, and software

packages to be delivered to Samsung.

References

[1] M. Hansard, R. Horaud, M. Amat, and S. Lee. Projective Alignment of Range and Parallax Data. In Proc. Computer Vision and Pattern

Recognition, pages 3089-3096, Colorado Springs, CO, June 2011.

[2] A. Zaharescu, E. Boyer, and R. P. Horaud. Topology-Adaptive Mesh Deformation for Surface Evolution, Morphing, and Multi-View

Reconstruction. IEEE Transactions on Pattern Analysis and Machine Intelligence, 33(4):823-837, April 2011.

Eligibility

 Applicants should have strong background in multiple-camera and stereo vision algorithms (calibration, rectification, 

 matching, reconstruction), knowledge in 3D geometric processing (mesh processing, photo-realistic rendering), and strong skills in numerical

methods and programming. Please provide a complete CV, the url of a publication list, as well as the names and emails of at least two

recommending persons.

Start date: January-March 2012

Contact person: Radu Patrice HORAUD <Radu.Horaud<στο>inria.fr>

Deadline: 31 December 2011