Accelerated volume rendering and tomographic reconstruction using texture mapping hardware

Abstract

Volume rendering and reconstruction centers around solving two related integral equations: a volume rendering integral (a generalized Radon transform) and a filtered back projection integral (the inverse Radon transform). Both of these equations are of the same mathematical form and can be dimensionally decomposed and approximated using Riemann sums over a series of resampled images. When viewed as a form of texture mapping and frame buffer accumulation, enormous hardware enabled performance acceleration is possible.