Noise suppression in cone beam CT projection data

A method for suppressing noise in a diagnostic 3-D image, executed at least in part on a logic processor, captures, at each of a number of projection angles, 2-D image projection data, wherein each 2-D image projection has a central pixel and arranges the 2-D image projection data to form a 3-D data set. Each of the 2-D image projections is processed by performing a diffusion filtering process that obtains a homogeneity value for the 3-D data set, generates a diffusion conductance function according to an intensity gradient between adjacent digital image elements, and applies the diffusion filtering process to digital image elements according to the obtained homogeneity value, the generated diffusion conductance function, and a weighting value that relates to the distance of each pixel in the projection from the central pixel. The diagnostic 3-D image is reconstructed from the processed 2-D image projections.