Making accurate measurements from radiographic images is time consuming and expensive. Simple and inexpensive diagnostic/monitoring systems will require a high degree of automation if they are to be used by radiographers or nurses in routine clinical practice without special training. Completely automatic image analysis systems are rare in radiology, although a number of experimental automated (initial human interaction) systems do exist. No fully automatic system exists for analysing dental bitewing radiographs. The role of multiple layer image models and multiple sources of information are discussed in relation to dental images. The potential role of interdental spaces is suggested as necessary key image regions to be identified for the model building process. A theoretical image model data structure is proposed which is experimentally tested in the following paper.