Scattering correction and image restoration in neutron radiography and computed tomography



Journal Title

Journal ISSN

Volume Title



Neutron radiography and computed tomography are nondestructive imaging techniques for the assessment of internal structure of objects. Neutron scattering in such objects can cause image degradation and complicate image interpretation. The removal of the scattering effect is one of the most challenging problems in neutron imaging. In this work, a new method for scattering correction is being developed. Experimental measurements and Monte Carlo simulations were used to investigate the effect of thermal and fast neutron scattering on neutron image degradation both qualitatively and quantitatively. Neutron scattering degrades the quality of radiographs and in cases of severe scattering, could blur sharp edges in neutron images. In addition, neutron scattering imposes an error in neutron radiography quantitative measurements. In this study, scattering correction was approached in two different ways. The first consisted of image restoration using the Slow Evolution from the Continuation Boundary (SECB) method. The SECB method was investigated and implemented to deblur neutron images for such cases when neutron scattering effect is severe enough to blur produced radiographs. The SECB method is a noniterative linear image deblurring method based on the slow evolution constraint, which is highly effective in suppressing noise amplification. The second approach for a scattering correction, which has been developed for the first time as part of this study, is based on the conjecture that there exists a correlation between the pattern of scattered neutrons as observed from a given side of the object as the object is irradiated from different sides. This suggests rotating the sample with some angle to clear the direct neutron view and obtain an image of pure scattering. The correlation between this side image and the scattering component of the forward image could be used to obtain an estimate of the forward scattering component. The estimated scattering component would then be subtracted from the degraded image to get a scattering-free image. Data manipulation of the scattering side-images was used to correlate the scattering side-image to the forward scattering component utilizing the scattering information outside of the neutron beam scope. Another approach was to implement artificial neural networks to capture the correlations between scattering side-images and the forward scattering components as obtained from numerical simulations for typical samples and utilize these networks to get an estimate of the forward scattering component for the object of interest.