Real-time diffraction computed tomography data reduction

Abstract

Diffraction imaging is an X-ray imaging method which uses the crystallinity information (cell parameter, orientation) as a signal to create an image pixel by pixel: a pencil beam is raster-scanned onto a sample and the (powder) diffraction signal is recorded by a large area detector. With the flux provided by third-generation synchrotrons and the speed of hybrid pixel detectors, the acquisition speed of these experiments is now limited by the transfer rate to the local storage as the data reduction can hardly be performed in real time. This contribution presents the benchmarking of a typical data analysis pipeline for a diffraction imaging experiment like the ones performed at ESRF ID15a and proposes some disruptive techniques to decode CIF binary format images using the computational power of graphics cards to be able to perform data reduction in real time.This paper describes benchmarking of the data analysis pipeline for X-ray diffraction computed tomography and CIF binary format byte-offset decompression on a GPU.