Wavelet-based fast segmentation of a tooth imaged with Digital Imaging Fiber Optic Transillumination (DIFOTITM)

The presented segmentation algorithm is computationally cost-effective and allows one to segment a tooth of interest from DIFOTI images solely in the wavelet transform domain, i.e. no inverse wavelet transform is required. The algorithm exploits the characteristics of the absorption and scattering of photons through the tooth and makes use of “redundant” wavelet representations of the DIFOTI imagery. The results show that all the teeth in the test sets were successfully segmented and that the time required to detect the boundary of a tooth scales linearly [O(n)] with the total number of pixels. Cases of DIFOTI images for which the method requires exception handling are identified, and the resolutions employed for these cases are demonstrated