Dental Identification based on Teeth and Dental Works Matching for Bitewing Radiographs

This paper presents an enhanced human identification method based on matching both the contours of teeth and the shapes of dental works (DWs) using bitewing radiographs. To reduce teeth matching error due to unsatisfactory alignment of two incomplete tooth contours, we propose an enhanced contour alignment by pruning the outliers from both contours after they are aligned with the original contours using an adopted regular alignment method then realigning the pruned contours again, and a revised Hausdorff distance for measuring the distance between two contours. To compensate DW matching error based on the distance of DWs’ contours due to imperfect alignment of the teeth in which they reside, we propose using alignment-invariant Fourier descriptors of DWs’ contours as an additional feature. The distance between the contours of teeth and the distance between the shapes of DWs are weightily combined for improving identification matching. Experiments on a small database show that (1) matching with pruned tooth contours improved 8% accuracy for the top-1 retrieval when compared to the result of matching with original contours; (2) matching with the combined frequency and spatial features of DWs improved about 19% and 5% accuracy, respectively, when compared to matching with only the spatial and the frequency feature for the top-1 retrieval; and (3) matching with the pruned tooth contours and both the spatial and frequency features of DWs, the retrieval accuracy in identification matching achieved to 100% for the upper, the lower, and both jaws.