Description

Direct volume rendering of large volumetric data sets on programmable graphics hardware is often limited by the amount of available graphics memory and the bandwidth from main memory to graphics memory. To avoid this problem often the data is given in the form of unstructured grids where high frequency detail information is sampled with a higher number of points and so less points are required. In many fields such as Flow Visualisation, Data Simulation, Astronomical Data Visualisation the application of such rendering techniques is very important and is currently achieved through unstructured tetrahedra grid rendering. We want to give up from connectivity information and instead of that use Unstructured Points Sets efficiently to render and visualise this large datasets.

Task

The starting point of this topic will be understanding recent papers on Unstructured Point Set Rendering. Then your task will be one of the following: Implementing one or several of these ideas in a GPU-based Raycasting approach with the intention of extending them. Implementing a CUDA-based parallel solution of the variational reconstruction approach. Of course you can contribute your own ideas, rather than only implementing an algorithm or technique from a paper.

Requirements

Good knowledge and interest in computer graphics and visualization, very good programming skills.

Additional Information

Contact: Erald Vuçini

• Computer Graphics Group
• Courses
• Projekte
• Unstructured Point Sets