Performance verification of the "ghost"-free tomographic reconstruction method QURT

Abstract

We recently proposed a novel tomographic reconstruction method called the quantisation unit reconstruction technique (QURT) that reconstructs a tomogram by arranging grey-level quantisation units in three-dimensional image space through unique discrete processing. QURT significantly reduces artificial image “ghost” generation caused by the “missing-wedge” and the number of projection images, which are common limitations of conventional methods. Unlike other recently developed advanced methods, such as compressed sensing, QURT does not require any parameter adjustments or modeling based on prior knowledge but requires only a series of projection images and tilt-angle data. It does not select the processing target because it exploits the inherent constraints of projection theory. In this study, the practical performance and resolution of QURT were quantitatively evaluated in both real and frequency domains using Fourier ring correlation and other metrics in simulations with fine test patterns. Even in the presence of the missing-wedge, QURT generated almost no ghost image outside the test pattern region, considerably suppressing the artifacts. This advanced property was experimentally validated by applying HAADF-STEM tomography to analyse a catalyst sample (Pd/CeO₂–ZrO₂–Al₂O₃), which was particularly sensitive to ghost image generation inside pores. QURT clearly revealed the pore structure without compromising the low-contrast Al regions and without generating ghost images, even at a tilt step angle of up to 4°. In EDXS tomography, QURT also reconstructed a 3D elemental map even from a small number of 22 tilt series signal maps with the missing-wedges.