Implementation of 2D/3D Registration algorithms
using GPU hardware
|Thesis by:||Razvan Chisu|
|Advisor:||Prof. Dr. Nassir Navab|
|Supervision by:||Wolfgang Wein|
This project is offered in collaboration with Siemens Corporate Research (SCR) . The student will start the thesis at TUM and afterwards spend 3 month abroad at the SCR Lab in Princeton, USA.
Registering a two-dimensional X-Ray image with a volumetric CT data set is a computationally demanding task. It comprises repeated generation of simulated X-Ray images from the CT data, which are compared iteratively with the real X-Ray. This simulation, which is basically a volume rendering of the CT data, can be done more efficiently using the 3D texturing features available on modern graphics hardware such as NVidia Geforce FX boards. The main idea is to move even more of the computations to the graphics card. In the scope of this Diplomarbeit, the student will first understand the 3D/2D registration procedure, and then evaluates different options for porting image registration algorithms to the GPU. Implementations will be done using the Cg language for GPU programming.-- RazvanChisu - 08 Jul 2004