Statistical Comparison of Various Reconstruction Algorithms with respect to Missing Wedge Artifacts in Computed Tomograp
- PDF / 1,923,943 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 103 Downloads / 193 Views
Statistical Comparison of Various Reconstruction Algorithms with respect to Missing Wedge Artifacts in Computed Tomography Sebastian Lueck1, Andreas Kupsch2, Axel Lange2, Manfred P. Hentschel2 and Volker Schmidt1 1
Institute of Stochastics, Ulm University, Helmholtzstr. 18, 89081 Ulm, Germany BAM Federal Institute of Materials Research and Testing, Unter den Eichen 87, 12205 Berlin, Germany 2
ABSTRACT The presence of elongation, streak and blurring artifacts in tomograms recorded under a missing wedge of rotation angles presents a major challenge for the quantitative analysis of tomographic image data. We show that the missing wedge artifacts of standard reconstruction algorithms may be reduced by the innovative reconstruction technique DIRECTT. For the comparison of missing wedge artifacts we apply techniques from spatial statistics, which have been specifically designed to investigate the shape of phase boundaries in tomograms. INTRODUCTION The theory of tomographic reconstruction of a function f from its lower-dimensional projections assumes that these are available for all directions of a semicircle, thus forming the complete Radon transform of f. Nevertheless, in particular electron tomographic techniques such as STEM [1] usually suffer from an entire missing interval of rotation angles, the so-calledmissing wedge (MW). This is due to technical or physical limitations restricting the tilt range of a sample. Sometimes even in X-ray tomography large or strongly absorbing flat samples restrict the accessible angular sector. In this situation solutions to the tomographic reconstruction problem are no longer unique. If standard tomographic reconstruction algorithms such as filtered backprojection (FBP) [2] or SIRT [3] are applied to an incomplete Radon transform with a MW of rotation angles, reconstructions exhibit elongation, blurring, and streak artifacts (Fig.1). In particular elongation artifacts, which e.g. distort circular objects to lemon shapes, pose a major problem for quantitative image analysis of tomographic data. From an experimental point of view, suggestions have been made to reduce MW artifacts by combining two tilt series with orthogonal tilt axes leaving only a missing pyramidal wedge of projection information [4]. On the other hand, MW artifacts may also be reduced by advanced reconstruction algorithms. One approach is to deal with the missing uniqueness of solutions to the reconstruction problem by restricting the space of possible solutions using a priori information on the composition of the sample. This is done in discrete tomography where the number and possibly even the value of different grayscales in the reconstruction are parameters of the algorithm [5]. This way the algorithm computes a solution within the much smaller space of images with a limited number of gray values. The algorithm DIRECTT (Direct Iterative Reconstruction of Computed Tomography Trajectories) [6, 7], which is in focus of this study, can be used in an analogous or discrete tomography mode. However, the central princi
Data Loading...
