In a previous work we developed an algorithm to improve the Hounsfield Unit accuracy of CBCT images based on prior information from a CT scan. However, the processing time required to run the algorithm may be a barrier to clinical implementation. Here we describe work to speed up two key processing steps: 3D binary morphological operations and image interpolation. Efficient binary morphological operators have been implemented in three dimensions using C++, extending the open source leptonica image processing library. Processing time comparisons have been made to implementations of three dimensional binary morphology available in ITK, MATLAB and IDL. The efficient implementations presented in this report have been found to require processing times up to three orders of magnitude shorter than the alternatives. Image downsampling has also been investigated as a method to enable faster processing. Downsampling images by a factor of 2.0 (4.0) can produce a speedup of 1.8x (3.4x) in a processing step that interpolates into masked regions of an image. Preliminary studies of the effect of downsampling on the quality of final processed images suggest that downsampling by a factor of 2.0 produces a negligible decrease in final processedimage quality. When used together these two developments can allow significantly faster processing of CBCT images.