An inexpensive image processing system, assembled from commercially available components and driven by "in-house" software, was tested as a means of processing radiographs obtained in a radiology department. In this preliminary study, the main interest was to investigate, through subjective appraisal, the ability of the system to recover the image quality of poorly exposed radiographs. Subsidiary objectives were to obtain some information regarding the specification limits of the system and its components, as well as some quantitative estimate of its advantages for clinical diagnosis. The system was used to process a set of film radiographs of an anthropomorphic chest phantom, produced under conditions that were designed to provide a wide range of image qualities. Following this, four sets of randomly ordered processed and unprocessed images were formally assessed by a group of experienced observers. Consensus opinion from the group indicated that significant improvement could be obtained through simple image processing. Some radiographs regarded as unacceptable became, after processing, acceptable to viewers.