(closed) Master Project: Spatio-temporal fusion of range and stereo data for high-quality depth sequences

Short Description

High-quality depth sequences are required by lots of applications such as 3DTV and film industry. The current range cameras are based on time of flight (TOF) and provide either low-resolution (e.g. Mesa SR4000) or mid-resolution (e.g. Kinect 2) depth maps. Commonly, these cameras are used along with a stereoscopic pair of color cameras in order to fuse range and stereo data for obtaining high-resolution (HR) depth maps. However, the final HR depth map suffers from inaccuracies and missing data regardless of the fusion technique. Moreover, frame-wise processing makes all side-effects flicker when a depth sequence is required.

In this project, the development of a spatio-temporal fusion model of range and stereo data will be investigated. A multi-resolution spatio-temporal model that deals with noisy and incomplete HR maps from a frame-wise fusion technique and provides temporally coherent high-quality depth maps will be investigated.

The project is suitable for a second year master student with very good background in computer vision and optimization as well as very good programming skills in Matlab and C/C++. The project is part of a broader effort to build high-quality depth streams for 3DTV and is funded by the ANR project MIXCAM. 

The project may start anytime after 1 February 2015 for a period of six months.

Information for applicants: Please send your complete CV, university grades, and the names and emails of two recommending persons to both georgios.evangelidis@inria.fr and sileye.ba@inria.fr. Students enrolled in a French university will receive a monthly net salary of 430€. Students enrolled in another university will receive a monthly net salary of 1100€. Please note that some restrictions apply to non-French students and their admission is conditioned by an approval from the French Ministry of Defense.


[1] J.Sanchez-Riera, J. Cech, R. Horaud:  Robust Spatiotemporal Stereo for Dynamic Scenes, IEEE ICPR 2012

[2] J. Cech, J. Sanchez-Riera, R. Horaud: Scene flow estimation by Growing Correspondence Seeds, IEEE CVPR 2011

[3] S. Ba, T. Corpetti, R. Fablet: Multi-resolution missing data interpolation in SST image series, IEEE ICIP 2011

[4] G. Evangelidis, M. Hansard, R. Horaud: Fusion of Range and Stereo Data for High-Resolution Scene-Modeling, IEEE TPAMI 2015 (accepted)

[5] K. Ruhl, C. Lipski, M. Magnor: Integrating Approximate Depth Data into Dense Image Correspondence Estimation, CVMP 2012