IMAGE REGISTRATION-BASED APPROACH TO RANKING DENTAL X-RAY IMAGES FOR HUMAN FORENSIC IDENTIFICATION

Dental features have been widely used for forensic identification purposes. However, the point-by-point comparison performed by a forensic odontologist in mass disaster situations or missing persons cases would be cumbersome and time-consuming due to the large number of records that would need to be compared. Consequently, a move towards computer-aided dental identification systems is necessary. In this work, we propose a computer-aided framework for matching of dental radiographs based on image registration. Given a postmortem (PM) radiograph with a marked region of interest (ROI), we searched the database of antemortem (AM) radiographs to retrieve a closest match. To express the degree of similarity/overlap between two radiographs, we used the weighted sum of squared differences (SSD) cost function. The method was tested on a database of 571 radiographs belonging to 41 distinct individuals. In 90% of the identification trials, our method ranked the correct match in the top 10%. In all trials, the correct match was among the top 22%. These experiments indicated that matching dental records using the SSD cost function is a viable method for human dental identification.