Department of Computer Science 5

Contact

cs5-info(at)lists.fau.de

+49-9131-85-27775

+49-9131-85-27270

Secretary

Monday	8:00 - 12:15
Tuesday	8:00 - 16:45
Wednesday	8:00 - 16:45
Thursday	8:00 - 16:45
Friday	8:00 - 12:15

Address

Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU)
Lehrstuhl für Informatik 5 (Mustererkennung)
Martensstr. 3
91058 Erlangen
Germany

Powered by

Dept. of Computer Sc. » Pattern Recognition » Research » Former Projects » Hardware Acceleration Techniques for 3D Cone-Beam Reconstruction

Hardware Acceleration Techniques for 3D Cone-Beam Reconstruction

no description

The video on the right shows an on-the-fly (FBP) reconstruction and visualization of a head acquired by a C-arm CT system.

The reconstruction and visualization was performed on first generation CUDA-capable graphics cards by NVIDIA a GeForce 8800 GTX with 768MB memory. The reconstructed volume is a 512³-cube using floats (512MB). For the reconstruction 414 projections, each a size of 1024² floating-point pixels, were streamed onto the graphics card and than back-projected on the volume. Details about the implementation and reconstruction speed can be found in our paper. Note that the direkt visualization of a slices during reconstruction reduces the computational speed due to extra costs. After all projections are back-projected we slide through the volume to show that not only one slice was reconstructed.