Navigation with Plenoptic Imaging on
Virtual Reality Scenarios

 

 

Description

 

Light-field representations (LF images) are acquired by plenoptic or light-field cameras, which can be thought off as arrays of pinhole cameras. One array of cameras can be represented compactly by an intrinsic matrix [Dansereau13, Zhang18], like what happens with pinhole cameras. A moving camera is an alternative way to acquire light-field (LF) images [Mildenhall19].

 

Considering a set of visual landmarks, metric depth information can be obtained, provided the LF camera is calibrated. This is an interesting difference as compared, for example, to do monocular SLAM. One central objective of the thesis is assessing whether LF information effectively helps navigation and provides a metric nature to the estimated motion. A specific objective would be realizing a statistical study about how the LF cameras baseline effects on depth estimation accuracy.

 

Detailed steps:

- LF acquisition with a moving pinhole camera

- LF video acquisition and 3D Scene Flow estimation

- LF features integration in a SLAM framework

 

 

References:

 

[Dansereau13] Dansereau, Donald G., Oscar Pizarro, and Stefan B. Williams. "Decoding, calibration and rectification for lenselet-based plenoptic cameras." Proceedings of the IEEE conference on computer vision and pattern recognition. 2013.

 

[Zhang18] Zhang, Qi, et al. "A Generic Multi-Projection-Center Model and Calibration Method for Light Field Cameras." IEEE transactions on pattern analysis and machine intelligence (2018).

 

[Mildenhall19] Local light field fusion: Practical view synthesis with prescriptive sampling guidelines. Mildenhall B, Srinivasan PP, Ortiz-Cayon R, Kalantari NK, Ramamoorthi R, Ng R, Kar A. ACM Transactions on Graphics (TOG). 2019 Jul 12;38(4):1-4. https://github.com/Fyusion/LLFF

 

 

Related previous dissertations:

 

"Structure Reconstruction using Plenoptic Cameras", Simão Marto, MSc Thesis, MEAER, IST - 2016/2017,

https://fenix.tecnico.ulisboa.pt/cursos/meaer/dissertacao/1409728525632074

 

"Deep Depth from Plenoptic Images", Diogo Portela, MSc Thesis, MEEC, IST - 2017/2018,

https://fenix.tecnico.ulisboa.pt/cursos/meec/dissertacao/1972678479054436

 

"Plenoptic Face Reconstruction", Gonçalo Pedro, MSc Thesis, MEEC, IST - Oct.2021,

https://fenix.tecnico.ulisboa.pt/cursos/meec/dissertacao/846778572213414

 

"Visual Anticipation of Collisions with Moving Objects based on Plenoptic Cameras", Miguel Rodrigues, MSc Thesis, MEEC, IST - Nov.2021

 

"Animated Sequences from Image Based Rendering", António Maia, MSc Thesis, MEEC, IST - started Sep.2021

 

 

Place for conducting the work-proposal:

 

ISR / IST

 

 

 

More MSc dissertation proposals on Computer and Robot Vision in:

http://users.isr.tecnico.ulisboa.pt/~jag