Usability of Skin Texture Biometrics for Mixed-Resolution Images

Abstract

There is a growing demand for alternative biometric modalities that can handle various real world challenges such as recognizing partially occluded individuals. Skin texture has been proposed as a potential alternative; however, such skin texture analysis can become difficult when captured images are at varying resolutions (due to different distances or devices). This paper explores the prospect of using mixed-resolution facial skin images as a source of biometric information. The four facial skin regions investigated here are the forehead, right cheek, left cheek, and chin which were selected because at least one of these are expected to be captured in real-world scenarios. The proposed framework first localises and assesses the usability of the extracted region of interest (ROI) for subsequent analysis. Local Binary Pattern (LBP) descriptors are then used for feature matching because of their reported effectiveness in extracting skin texture information. Experiments conducted using the XM2VTS database suggest that mixed resolution skin texture images can provide adequate information for biometric applications.