Enhanced Patch-wise Maximal Intensity Prior for Deblurring Neutron Radiographic Images

Abstract

In neutron radiographic imaging, generally, the collimation ratio is assumed to be sufficiently large to ensure a valid approximation for parallel beam geometry. However, this assumption is difficult to apply in small nuclear reactors due to the low- intensity neutron flux. For this reason, these reactors produced inherently blurry neutron images. In this paper a blind deconvolution technique is investigated for the enhanced visual quality of neutron images through the reduction of blurring artefacts. Technically, this approach is extremely challenging because it requires an unknown point spread function. To solve this problem, scholars employ the gradient minimization strategy under the framework of a maximum a posterior, which leads to the development of an improved deblurring method, referred to in this paper as the enhanced patch-wise intensity prior. Experimental results demonstrate that the high competitiveness of the proposed method in terms of blind or no-reference evaluation measure, with an average of 46.1 for six neutron images used in this study. This value is considerably lower compared with those of existing deblurring techniques, which implies a more accurate restoration. Additionally, the proposed method resulted in the highest, and hence, the best entropy and contrast values, averaging at 7.09 and 1.05 respectively. The proposed method is also the second fastest technique witd mean time of 180 s.