Challenges in using GPUs for the reconstruction of digital hologram images

Abstract

In-line holographic imaging is used for small particulates, such as cloud or spray droplets, marine plankton, and alluvial sediments, and enables a true 3D object field to be recorded at high resolution over a considerable depth. To reconstruct a digital hologram a 2D FFT must be calculated for every depth slice desired in the replayed image volume. A typical in-line hologram of ∼ 100 micrometre-sized particles over a depth of a few hundred millimetres will require O(1000) 2D FFT operations to be performed on an hologram of typically a few million pixels. In previous work we have reported on our experiences with reconstruction on a computational grid. In this paper we discuss the technical challenges in making efficient use of the NVIDIA Tesla and Fermi GPU systems and show how our reconstruction code was optimised for near real-time video slice reconstruction with holograms as large as 4K by 4K pixels. We also consider the implications for grid and cloud computing approaches to hologram replay, and the extent to which a GPU can replace these approaches, when the important step of locating focussed objects within a reconstructed volume is included.