Beam hardening correction for X-ray computed tomography of heterogeneous natural materials

We present a new method for correcting beam hardening artifacts in polychromatic X-ray CT data. On most industrial CT systems, software beam-hardening correction employs some variety of linearization, which attempts to transform the polychromatic attenuation data into its monochromatic equivalent prior to image reconstruction. However, determining optimal coefficients for the transform equation is not straightforward, especially if the material is not well known or characterized, as is the usual case when imaging geological materials. Our method uses an iterative optimization algorithm to find a generalized spline-interpolated transform that minimizes artifacts as defined by an expert user. This generality accesses a richer set of linearization functions that may better accommodate the effects of multiple materials in heterogeneous samples. When multiple materials are present in the scan field, there is no single optimal correction, and the solution can vary depending on which aspects of the beam-hardening and other image artifacts the user wants to minimize. For example, the correction can be optimized to maximize the fidelity of the object outline for solid model creation rather than simply to minimize variation of CT numbers within the material. We demonstrate our method on a range of specimens of varying difficulty and complexity, with consistently positive results.