An adaptive algorithm for gamma spots removal in CCD-based neutron radiography and tomography

In thermal neutron radiography and tomography, a lot of gamma spots may appear on CCD images as the result of the interaction between gamma photons and CCD pixels. These bright spots may have very high magnitudes and may involve more than 20 pixels each. They are harmful to quantitative data analysis and 3D reconstruction, and are difficult to be removed without loss of image quality. Some state-of-the-art edge preserving algorithms, such as bilateral total variation (BTV) filter and non-local means filters, cannot remove these spots because of their big sizes and high magnitudes. We report a new intuitive algorithm working on a ‘find and replace’ strategy. A Laplacian of Gaussian (LoG) filter is applied to the gamma spots contaminated raw data to diminish its constant components and amplify only the fast varying contents including edges and gamma spots, then the LoG image is segmented with adaptive thresholds to locate the gamma spots and form a substitution map. Spot substitution is performed according to the map with the median of its neighborhood, the size of which is chosen according to its magnitude. This method can achieve satisfactory spots removal with better preservation of image details compared with the now widely used median filters. Results from this algorithm and comparison with other methods are presented.