Fast interactive programs for the investigation of the effects of different processing methods were developed in an attempt to optimize the diagnostic value of digital scanned projection radiography (SPR) data. Filters from two classes of functions with varied frequency characteristics were evaluated. Arbitrary combinations of filtered and unfiltered images were generated. Speed was achieved by implementing the filtering as a convolution operation with standard computed tomography hardware (20 sec/image). The criteria for the choice of the reconstruction parameters are discussed. The clinical results show that suppressing the low-frequency content of the image, as in smoothing, low-pass filtering, and subtracting the background from the original image, optimizes the diagnostic value of SPR images.