DART: A robust algorithm for fast reconstruction of three-dimensional grain maps

摘要

A novel algorithm is introduced for fast and nondestructive reconstruction of grain maps from X-ray diffraction data. The discrete algebraic reconstruction technique (DART) takes advantage of the intrinsic discrete nature of grain maps, while being based on iterative algebraic methods known from classical tomography. To test the properties of the algorithm, three-dimensional X-ray diffraction microscopy data are simulated and reconstructed with DART as well as by a conventional iterative technique, namely SIRT (simultaneous iterative reconstruction technique). For 100 100 pixel reconstructions and moderate noise levels, DART is shown to generate essentially perfect two-dimensional grain maps for as few as three projections per grain with running times on a PC in the range of less than a second. This is seen as opening up the possibility for fast reconstructions in connection with in situ studies.