Offers listWhy France ?F.A.Q.Email AlertsPost an offer
or

Forgot Password ?
Offer details


Light-field remote vision

Full Description:

Recently, light-field cameras arrived on the consumer market in the form of the Lytro camera (Raytrix also sells a light-field camera more targeted on the industry and research), in which a microlens array has been inserted between the main lens and sensor plane allows refocussing the photograph off-line. However, there exists a large variety of light-field capturing devices, from microscopes [7], to large sets of regularly-arranged cameras such as the Stanford multi-camera array [4], to unstructured sets of cameras looking roughly in the same direction [8,9].

Conventional cameras follow the pinhole camera model inherited from the biological eye: the image records the radiance of the 2-dimensional set of optical rays that pass through a single point in space, the optical center. The light-field camera [1,2] samples what is called the plenoptic function, which contains radiance information about the set of 3D rays in space (which is 4-dimensional). This camera is usually built by placing an array containing thousands of lenslets between the main lens and the image plane. The result is somewhat equivalent (depending of the respective position of the lens array and the image plane [3]) to a 2D array of thousands of very low-resolution cameras (for example, the Lytro raw image is 3280x3280, and each lenslet occupies about 10x10 pixels).

Light-field imaging offers new ways of visualizing the appearance of a scene. Previous work showed that the captured images can be digitally refocussed [2], partially occluded areas can be made fully visible by combining the information from all viewpoints, giving the ability to "see through" partial occluders such as vegetation [4], and the 3-D geometry of the scene can be reconstructed from the light-field with high accuracy [10].

In this project, we propose to work specifically on long-range light-field imaging, where the acquisition device is used to visualize a scene which is typically a few meters in diameter, seen from a few hundred meters away, resulting in a form of Light-field telescope. The microlens-based plenoptic cameras are not well suited for this situation, because the subset of the plenoptic space sampled at this distance would be too small for practical applications. We thus propose to work on evaluating and optimizing different camera arrangements for various tasks (refocussing, 3-D reconstruction, stereoscopic 3-D visualization using novel-view synthesis [5]), in order to obtain the best trade-off between data volume, practical camera placement, and image quality [6]. The studied camera arrangements will include traditional setups such as a linear array of cameras or a 2-D array of cameras. However, more exotic setups may have advantages in the case of long-range vision, such as a constellations of small camera groups (which combine different sampling resolutions of the light field rather than having a fixed resoluytion), and these original combinations have not been studied before.

Specific problems due to long-range imaging will most probably arise in this situation, such as clear-air refractive-index perturbations due to temperature gradients, transmission index variations due to haze or smoke, and low-light due to the small solid angle in which the scene is contained (as seen from the acquisition device). The image degradation caused by these issues will be numerically quantified, either by simulation or by analysis, and possible solutions will be proposed, by combining the information from all cameras. For example, low-light could be compensated by considering the fact that the sum of camera apertures in the system is much larger than this of a single camera.

Application process

This PhD position is availaible to EU citizen only, and is co-funded by the French Government Defence Agency (DGA).

Send a motivation letter and a resumé (CV) by email to frederic.devernay at inria.fr with the subject line "[PhD] Light-field remote vision".

Bibliography

[1] Adelson, E. H., and Wang, J. Y. A. 1992. Single lens stereo with a plenoptic camera. IEEE Trans. Pattern Anal. Mach. Intell. 14, 2, 99.106, doi:10.1109/34.121783.

[2] Ng, R., Levoy, M., Bredif, M., Duval, G., Horowitz, M., and Hanrahan, P. (2005). Light field photography with a hand-held plenoptic camera. Stanford University Computer Science Tech Report CSTR 2005-02.

[3] Georgiev, T., Zheng, K. C., Curless, B., Salesin, D., Nayar, S., and Intwala, C. 2006. Spatio-angular resolution tradeoffs in integral photography. In Rendering Techniques 2006: 17th Eurographics Workshop on Rendering, 263-272, doi:10.2312/EGWR/EGSR06/263-272.

[4] Vaibhav Vaish, Richard Szeliski, C.L. Zitnick, Sing Bing Kang, Marc Levoy, Reconstructing Occluded Surfaces using Synthetic Apertures: Stereo, Focus and Robust Measures; Proc. IEEE CVPR 2006, doi:10.1109/CVPR.2006.244.

[5] Frédéric Devernay and Adrian Ramos Peon. 2010. Novel view synthesis for stereoscopic cinema: detecting and removing artifacts. In Proceedings of the 1st international workshop on 3D video processing (3DVP '10). ACM, New York, NY, USA, 25-30, doi:10.1145/1877791.1877798.

[6] A. Levin, W. T. Freeman, F. Durand. Understanding camera trade-offs through a Bayesian analysis of light field projections. Proc. of the European Conference on Computer Vision (ECCV), Marseille, France, Oct 2008, doi:10.1007/978-3-540-88693-8_7.

[7] Marc Levoy, Ren Ng, Andrew Adams, Matthew Footer, Mark Horowitz, Light field microscopy, Proceedings of ACM SIGGRAPH 2006, doi:10.1145/1141911.1141976

[8] Chris Buehler, Michael Bosse, Leonard McMillan, Steven Gortler, Michael Cohen, Unstructured Lumigraph rendering, in Proc. SIGGRAPH 2001, doi:10.1145/383259.383309

[9] Abe Davis, Marc Levoy, Fredo Durand, Unstructured Light Fields, in Proc. EUROGRAPHICS 2012, doi:10.1111/j.1467-8659.2012.03009.x

[10] S. Wanner, B. Goldluecke, Globally Consistent Depth Labeling of 4D Lightfields, in Proc. IEEE CVPR 2012, doi:10.1109/CVPR.2012.6247656.



Start Date: 01 October 2014duration: 36 months, deadline to apply: 01 April 2014

Subject1:Computer Science and IT Subject2:Optical Physics 
Institution:INRIA City:Grenoble Region:Rhône-Alpes
Salary:20-25k€ annual gross Fund category:Not Defined Doct. School:Mathematics, science and technology of information, informatics

Please connect to your account to see contact details



Back to offer list

Back to home page


PhD in France Website.
© French Embassy in India, CampusFrance Inde, 2014.

Contact us

Related Links