The advantages of iterative reconstruction methods for high resolution X-ray tomography

With the aid of X-ray Computed Tomography (CT) the internal structure of complex objects can be reconstructed as a virtual 3D volume. Although iterative reconstruction algorithms such as the Simultaneous Algebraic Reconstruction Technique (SART) [1] are gaining popularity, algorithms based on filtered back projection, such as the algorithm of Feldkamp, David and Kress (FDK) [2], are still most commonly used for the reconstruction of CT-projection data. A drawback of iterative algorithms is that they increase the reconstruction time, even when efficient implementation on GPU is used. Although these algorithms should not be considered as the optimal choice in every case, there are several situations where they increase image quality in comparison with analytical algorithms. The FDK algorithm typically requires a complete dataset, including a large number of equiangular projections at a relatively long exposure time. It has been demonstrated that iterative reconstruction algorithms provide better noise handling and that they can increase image quality in case the total number of available projections is limited or when the projection data is limited to a certain angular range (limited angle tomography) [3]. A major advantage of iterative reconstruction is that the algorithms can be adapted so prior knowledge about the sample or the beam can be incorporated. This allows for the reduction of artifacts, for example resulting from beam hardening or the presence of metals in the investigated sample [4]. Additionally it can be used to reduce the number of required projections in specific cases, for example when a sample consists out of a discrete number of materials. The latter case is illustrated below, where the the Discrete Algebraic Reconstruction Technique for Experimental data (EDART) [5] is compared with a standard SART reconstruction of an aluminum foam sample with 16 projection images