Analysing Facial Regions for Face Recognition Using Forensic Protocols

This paper focuses on the analysis of automatic facial regions extraction for face recognition applications. Traditional face recognition systems compare just full face images in order to estimate the identity, here different facial areas of face images obtained from both uncontrolled and controlled environments are extracted from a person image. In this work, we study and compare the discriminative capabilities of 15 facial regions considered in forensic practice such as full face, nose, eye, eyebrow, mouth, etc. This study is of interest to biometrics because a more robust generalpurpose face recognition system can be built by fusing the similarity scores obtained from the comparison of different individual parts of the face. To analyse the discriminative power of each facial region, we have randomly defined three population subsets of 200 European subjects (male, female and mixed) from MORPH database. First facial landmarks are automatically located, checked and corrected and then 15 forensic facial regions are extracted and considered for the study. In all cases, the performance of the full face (faceISOV region) is higher than the one achieved for the rest of facial regions. It is very interesting to note that the nose region has a very significant discrimination efficiency by itself and similar to the full face performance.