Robust Sclera Segmentation for Skin-tone Agnostic Face Image Quality Assessment

Face image quality assessment refers to the process of evaluating the utility of a face image for face recognition. It involves analyzing various quality factors that may impact the recognition performance. The quality measures produced from analyzing the image can be in the form of individual quality components, such as background uniformity, illumination uniformity, pose, exposure, dynamic range, sharpness, facial expressions, or in the form of a unified quality score. The ISO/IEC CD on 29794-5 [IS] (Information technology -- Biometric sample quality -- Part 5: Face image data) specifies that a face image quality assessment algorithm should be insensitive to demographic factors such as age, skin-tone or ethnicity. The eye sclera refers to the outer layer of the eyeball surrounding the iris. It is the opaque, whitish portion of the eye that surrounds the colored iris and the dark circular opening called the pupil. Figure 1 illustrates the anatomy of the eye including the sclera. This characteristic of being whitish in color regardless of age, ethnicity and skin-tone [Ka23] is what makes it interesting for the task of face image quality assessment. Analyzing the eye sclera in a face image can help in making the quality assessment algorithms of some of the face image quality components invariant to skin-tone and ethnicity.