Integrating a Profile of Frontal Face With Its Mirror Image for Facial Reconstruction.

Abstract

Face recognition plays an important role in the field of biometrics, bioinformatics, and forensic identification. Human faces appear bilaterally symmetric. However, asymmetry is a norm in the human body and careful morphological and metric examination of the human face gives an exact idea of the asymmetry present in the face of an individual. Most of the studies on asymmetry of human face have been carried out with regard to evaluation of expressions in psychology, anthropology, and biometrics. In this preliminary study, an idea of the human face symmetry has been used for complete facial reconstruction. Sometimes, in the CCTV cameras and other surveillance systems, the complete face is not visible, but only a half of the frontal face is detectable. In all such cases, the mirror image of the face can be used to reconstruct the complete face of the person in question for representation-based face recognition. In previous studies related to forensic identification and in the discipline of computer science, mirror images have only been used to overcome and deal with the problem of nonsufficient training sample and have not been used for the reconstruction of the full facial profile. In this article, we present a computerized methodology to generate the mirror image from one frontal half of the face (either left or right) and merging the both (input image and generated mirror image) halves to create the full face. The program was designed using METLAB computer program. This procedure is very simple, less time-consuming, and computationally efficient. The results have been presented with regard to generate a complete face in those cases where only one frontal profile, that is, left or right sides of the face is available for examination. The scenario is commonly seen in those cases also where shadow almost hides one half of the face. The proposed methodology would also be useful for improving other facial reconstruction and recognition methods in forensic identification and in other medical fields such as plastic surgery, where creating full face still poses a technical challenge to a plastic surgeon. Some challenges, limitations, and future perspectives of the method have also been discussed in the article.