X-ray computed tomography relies on precise line-integral measurements of primary X-rays’ attenuation to reconstruct local material volumes (tomo-gram). However, primary X-rays can be deflected after interacting with the materials of an object (scatter), introducing severe quantitative inaccuracies in reconstructed tomograms, such as contrast loss.

While beam-stop array (BSA) method can directly sample object-scattered photons sparsely, the accuracy of this technique is compromised by several confounding factors: Scatter within the X-ray source and the cabinet wall of the CT machine, detector lag, where signal from previous exposures persists into subsequent projection frames, and the scatter within the detector. Un-accounted for, these confounding factors distort the sparsely sampled scatter signal and reduce the accuracy of object-scatter estimates.

Here, we present an experimental framework that systematically isolates and mitigates confounding factors prior to object-scatter estimation. Source and cabinet scatter is suppressed via source-collimating apertures, while detector scatter is corrected through deconvolution using a knife-edge-measured point spread function. With these secondary backgrounds eliminated, residual BSA shadow intensity strictly reflects object scatter, allowing accurate 2D interpolation across the projection domain. Our method was validated on an aluminium 6061 alloy extrusion with a titanium rod stick in the middle. This method restores quantitative attenuation uniformity, the geometrical distortion and it eliminates background air artifacts.