Exploring the potential of facial skin regions for the provision of identity information

This work presents a novel framework to investigate the possibility of using textural information from facial skin regions for biometric identification of individuals. Such information is of practical use when the entire facial image is not available for identifying individuals, such as in forensic investigations. Here, four facial regions were investigated (forehead, right cheek, left cheek, and chin), since they are likely to be visible even in partially occluded facial images and they are relatively easy to distinguish. Facial landmarks are automatically detected to facilitate the extraction of these facial regions of interest. A skin detection technique is applied to identify regions with significant skin content. Each skin region is then processed independently using features based on Local Binary Patterns and Gabor wavelet filters. Feature fusion of sub-regions is used prior to classification of the images. Experiments were carried out, using the publicly available Skin Segmentation database and the XM2VTS databases, to evaluate the performance of skin detection technique and biometric recognition respectively. The results show that the skin detection algorithm provided comparable results to other state-of-the-art techniques. Of the regions investigated, the forehead and chin regions were found to provide a rich source of biometric information.