Human forensic identification with dental radiographs using similarity and distance metrics

Abstract

Forensic dentistry involves the identification of people based on their dental records, mainly available as radiographic images. Human identification is done by matching the given post-mortem radiographs with the ante-mortem images. In this paper, a computer-aided dental identification system for matching dental records is presented. The tooth contour is used as a feature for matching here. The proposed algorithm consists of five stages. As an initial step, the image is preprocessed. Then it is isolated as individual tooth by radiograph segmentation. Shape extraction is made using connected component labeling. Finding similarity metric is the next step. Various distance measures are also applied to find better matching. Finally candidate matching is done by obtaining the percentage of match between the original and extracted shape using both the similarity and distance measures. Experimental results show that the finer matching distance is observed by the distance metric rather than similarity measures. The experimental results are obtained on a database of 100 dental images which includes both periapical and bitewing. The results show that a high hit rate is observed for the Euclidean distance measure and which is comparable with the other methods.