Imaging stitching is a solution for radiography and computed tomography (CT) applications where the object is larger than the beam size. Imaging stitching algorithms require a robust noise filter that maintains the landmark features used in stitching. In lens-coupled neutron radiography and CT, a camera is placed away from the neutron beam. Even with shielding, the camera experiences a high radiation dose of mixed gammas and neutrons. The CCD silicon sensor, sensitive to both gammas and neutrons, introduces speckled noise, pixel oversaturation, and blooming effects. Conventional median filters prove inadequate with this type of noise and can result in blurred images. Manual filtering of CT sets is timeconsuming and error-prone. An improved image filtering method designed for neutron CT data sets is therefore needed to improve imaging stitching algorithms. We have developed a method that utilizes statistical information in radiographs and variable-sized radii filtration to adequately remove noise while preserving resolution. Once noise has been identified, the algorithm tracks cluster size to inform local filter needs. Filtered radiographs are stitched using a semi-automatic algorithm. This approach works best for data containing features for joint corner detection. It does require specific user inputs, such as object size, features of interest, and alignment, to pinpoint the optimal joining location. Overall, our method represents a significant advancement in neutron CT image processing, offering improved results for imaging stitching and traditional CT applications. We describe the application of this combined filter and stitching algorithm on thermal and fast neutron CT data at OSURR.