State-of-the-art GPU rendering

Question

What is the current technique for implementing GPU-based volume rendering in a medical context, that is, reasonable amounts of image data (512x512x1500 or more per 16 bit) and interactive / real performance with optimal image quality? Is this still a racist approach presented by Kruger and Westermann (color cube, etc.)? Or did CUDA / OpenCL change the traditional Cg / GLSL game?

Thanks for any tips, links to papers, etc.! Matthias

Answer 1

Since then, nothing has changed. There are different versions of the CUDA / OpenCL / DirectCompute implementation, but they basically do the same thing.

Here's a good SIGGRAPH 2010 course. Volumetric methods in visual effects that generalize general methods.