Real-Time Radon Transform via the GPU Graphics Pipeline

Graphics processing units (GPUs) found in consumer hardware have emerged as first choice for computational-intensive image processing tasks, e.g., real-time computer tomography (CT) reconstruction. However, the GPU programming paradigm is rather intricate. To leverage the potential of GPUs for a broad audience, this paper is concerned with a self-contained, pedagogical implementation of the prototypical forward and inverse Radon transform. The full source code is available for download. Instead of utilizing the GPU merely as multiprocessor, the algorithm explicitly employs the graphics pipeline to take advantage of the dedicated GPU hardware components.