An improved algorithm for craniofacial reconstruction based on landmarks registration

Abstract

Craniofacial reconstruction is aiming at estimating and recovering the appearance of an unknown or unidentified skull, which is very useful in forensic detection. Traditional approach is tedious and subject, because experts stick the clay or plastic on the unidentified skull according to experience. So here we proposed a computer aided approach to reconstruct the outlook. Firstly, the 3D skull model is digitized through 3D laser scanner and the noise of the model is eliminated based on bilateral denoise. Secondly, an efficient landmark marking algorithm is proposed, meanwhile in order to make sure the accuracy of the landmarks, we can adjust the position of landmarks. Meanwhile an efficient approach of finding correspondence points is proposed, the preciseness of the landmarks will influence the final result. Finally, Thin-Plate spline (TPS) is adopted to achieve non-rigid registration between the unidentified skull and reference skull. Using the warps transformation from the reference skull to unidentified skull, the appearance of the unidentified skull can be estimated. The feasibility and usability of the algorithm are confirmed through related experiment.