Recognizing invariant geometric structure in dental radiographs.

The dental radiograph is a non-invasive tool that is used to view internal structures for the diagnosis of caries, periapical lesions and periodontal disease. The requirement for a standardized image is most prominent in periodontal disease since the diagnosis is best done with a difference radiograph. The difficulty is that exact registration for subtraction requires exact reproduction of imaging geometry. A new model of imaging geometry, based on the correspondence of 3D structures, to describe the radiograph formation process is presented. The experimental results show that 3D measurements can be made in dental radiographs (P < 0.01) with up to 16-mm translation errors and angulation errors of up to 32 degrees.