Abstract
In recent years phase-contrast has become a much investigated modality in radiographic imaging. The radiographic setups employed in phase-contrast imaging are typically rather costly and complex, e.g. high performance Talbot-Laue interferometers operated at synchrotron light sources. In-line phase-contrast imaging states the most pedestrian approach towards phase-contrast enhancement. Utilizing small angle deflection within the imaged sample and the entailed interference of the deflected and un-deflected beam during spatial propagation, in-line phase-contrast imaging only requires a well collimated X-ray source with a high contrast & high resolution detector. Employing high magnification the above conditions are intrinsically fulfilled in cone-beam micro-tomography. As opposed of 2D imaging, where contrast enhancement is generally considered beneficial, in tomographic modalities the in-line phase-contrast effect can be quite a nuisance since it renders the inverse problem posed by tomographic reconstruction inconsistent, thus causing reconstruction artifacts. We present an experimentally enhanced model-based approach to disentangle absorption and in-line phase-contrast. The approach employs comparison of transmission data to a system model computed iteratively on-line. By comparison of the forward model to absorption data acquired in continuous rotation strong local deviations of the data residual are successively identified as likely candidates for in-line phase-contrast. By inducing minimal vibrations (few mrad) to the sample around the peaks of such deviations the transmission signal can be decomposed into a constant absorptive fraction and an oscillating signal caused by phase-contrast which again allows to generate separate maps for absorption and phase-contrast. The contributions of phase-contrast and the corresponding artifacts are subsequently removed from the tomographic dataset. In principle, if a 3D handling of the sample is available, this method also allows to track discontinuities throughout the volume and therefore states a powerful tool in 3D defectoscopy.