3D data Representation Conversion



The recent progress in 3D data acquisition technologies enables live capturing sessions with cheap sensors. With this leap in data scanning, comes an urgent need for new data structures that accelerate data processing to real-time rates, in order to catch up with the high acquisition speeds.


We implemented a hierachical view-aligned data structure, which stores multi levels of depth intervals along pixels of the view plane. The next step is to fit the structure in the data acquisition-processing pipeline in order to conduct real life experiments and user studies. The Infinitam software (http://www.robots.ox.ac.uk/~victor/infinitam/) stores data from a Kinect camera in a volumetric structure which employs the voxel block hashing concept [1]. We need to transfer this volumetric data into our depth structure.

[1] K¨ahler, O., Prisacariu, V.A., Ren, C.Y., Sun, X., Torr, P.H., Murray, D.W.: Very high frame rate volumetric integration of depth images on mobile devices. IEEE Transactions on Visualization and Computer Graphics (Proceedings International Symposium on Mixed and Augmented Reality 2015) 21(11) (2015) 1241–1250


  • Get familiar with the infinitam software and its data representations, and be able to use it to capture data with a Kinect camera.
  • Get familiar with the implemented depth structure.
  • Convert the volumetric data into the depth structure format.
  • Live update the depth structure content with new frames data.


  • C++ programming skills.
  • Experience in openGL and/or CUDA [will help you finish faster]
  • Experience in point clouds processing is a plus (such as visualization,
    subsampling, normal estimation, ...etc).


Platform-independent C++.


For more information please contact Mohamed Radwan (radwan@cg.tuwien.ac.at) or Michael Wimmer (wimmer@cg.tuwien.ac.at).