Automated Skin Region Quality Assessment for Texture-based Biometrics

Designing a biometric system based solely on skin texture is of interest because the face is sometimes occluded by hair or artefacts in many real-world contexts. This work presents a novel framework for the assessment of skin-based biometric systems incorporating skin quality information. The quality or purity of the extracted skin region is automatically established using pixel colour models prior to biometric processing. Facial landmarks are detected to facilitate automated extraction of facial regions of interest. Although the present study is confined to the forehead region, the idea can be extended to other skin regions. Local Binary Patterns (LBP) and Gabor wavelet filters are utilised to extract skin features. Using the publicly available XM2VTS database, the experimental results show that the system provides promising performance when compared to other commonly used techniques.