Alternative representations for processing and visualizing large captured data sets



In collaboration with CRS4, the fellow will investigate novel representation forms for large data sets coming for example from laser scanners or multi-view stereo. In particular, for processing and visualizing these data, statistical representations should be investigated, for example based on Gaussian Mixtures, which were first introduced in this context by TUW at SIGGRAPH 2014. The goal of these representations is to find a reduced form of representation that can model the noise inherent in captured datasets using statistical properties, while taking up much less memory and allowing for much faster (i.e., interactive) processing than the original data set. The candidate will investigate different operators on such representations (filtering, smoothing, compression, etc.) and also investigate extensions from purely geometric data to color data and other attributes.

Expected Results

Research reports and internal prototype implementation, as well as published papers on statistical data representation for 3D data sets.

Specific Requirements

The position requires prior knowlede in computer graphics, computer vision, or visual computing. Hence, the background needs to be a M.Sc. in Computer Science or a closely related field. Skills in programming (C/C++, Matlab, OpenGL) are required.

Host Institution

Supervisor: Prof. Michael Wimmer
Head of Rendering and Modeling Group Institute of Visual Computing & Human-Centered Technology
TU Wien

Apply Here