Registration of Three Dimensional Human Face Images across Pose and Their Applications in Digital Forensic

Abstract

In digital forensic, three-dimensional face recognition is very challenging problem. The problem, with two-dimensional face recognition, is that, pose variation, illumination changes and expressions tend to reduce the face recognition rate. The problem aggravates in case of pose variations, especially when, the poses are rendered across extreme variations e.g. across 90 degrees. In contrast to two dimensional images, three dimensional images tend to reduce the shortcomings of two dimensional approaches. Since three dimensional images works with depth information, they do not depend on illumination, thus making the facial recognition system more robust. Pose variation affects three dimensional recognition rate hence, for a face to be recognized; it should be perfectly registered in three dimensional framework. In this book chapter, a comparative analysis of registration methods is presented for face recognition across different poses from 0 to 90°. Also, various registration approaches that are able to generalize identity, illumination and can also handle a given set of poses have been discussed in later sections. Also, several approaches used in the field of three dimensional face registration and recognition and their importance in digital forensics have been discussed. The application area of 3D faces recognition, in digital forensic, lies with the fact that, if the face of a person is oriented across a certain angle, he cannot be recognised in that position. So, perfect registration is very necessary to reconstruct the face, to be correctly recognized especially, for identifying subjects required for forensic study. Nowadays, photographic expert members from crime Investigation Bureau are researching for a number of new technologies such as facial recognition software, 3D modelling for crime scenes etc because most CCTV footages have very low picture quality which may not be much helpful for investigations. 3D face recognition system would be beneficial because they would help one to visualize the entire system, in all possible orientations.