Accelerating non-linear image registration with GPUs
Edinburgh Parallel Computing Centre, The University of Edinburgh
The University of Edinburgh, 2011
@article{ross2011accelerating,
title={Accelerating non-linear image registration with GPUs},
author={Ross, P.},
year={2011}
}
The alignment or registration of two images or volumetric datasets is frequently a requirement in modern image-processing applications, particularly within the context of medical imaging. Modern graphics-processing units (GPUs) are designed to perform simple 3D graphics-pipeline tasks on a massively parallel scale; this processing power can be harnessed for general computation via libraries such as Nvidia’s CUDA or the cross-platform standard OpenCL. By exploiting the unique hardware features of GPUs, a signi?cant performance improvement for registration applications can be achieved. As a result, the performance of one of the major bottlenecks has been improved by up to a factor of approximately 1000; this factor is likely to increase with larger datasets.
December 12, 2011 by hgpu
