The Influence of Data Filtering on Dimensional Measurements with CT

Abstract

Data processing is an essential part of the evaluation of measurement results, e.g. in X-ray computed tomography (CT). Some of these calculations are denoted as filtering. Filtering is always a trade-off between an enhancement of data quality and a loss of information. The standard reconstruction technique for CT data is t he filtered back-projection. Additionally to the high-pass filter inherent to this technique, a suppression of high-frequency noise is in most cases also necessary. Many different possible filter algorithms are found in the literature. Although their effect on the signal-to-noise ratio and the spatial resolution is well-studied, it is not obvious how the different filters affect the determination of the object's surface in the reconstructed data and, thus, the results of dimensional CT measurements. We compare the results of industrial dimensional CT measurements obtained at the same data set of a calibrated reference standard – a calotte cube –when using different filters and discuss the implications. Filters under investigation are mainly median filters on projections and as well filters during filtered backprojection and filters on the reconstructed volume.