Digital imaging fiber-optic transillumination (DI-FOTI) is a novel method to detect and monitor dental caries, using light, a charge-coupled device (CCD) camera, and computer-controlled image acquisition. The advantages of DIFOTI over radiography includeno ionizing radiation, no film, real-time diagnosis, and higher sensitivity in detection of early lesions not apparent to X-ray, as demonstrated in vitro. Here, we present a method of processing DIFOTI images, acquired at different times, for monitoring changes. Of central importance to this application is pattern matching of image frames that is invariant to translation and rotation of a tooth, relative to the field of view of the imaging camera, and that is robust to changes in illumination source intensity.

Our method employs:1) wavelet modulus maxima representations for segmentation of teeth images; 2) first and second moments of gray level representations of DIFOTI images in the spatial domain, to estimate tooth location and orientation; and 3) multiresolution wavelet magnitude representations for quantitative monitoring. Even with illumination source intensity variation, it is demonstrated in vitro that such wavelet representations can facilitate detection of simulated clinical changes in light transmission that cannot be detected in the spatial domain.