A single image based spectacle problem removal for improved ocular feature detection under spectacles

Ocular image processing is an important step in applications like face recognition, driver fatigue detection, iris recognition, and gaze tracking etc. Secondary reflections and glare formed on the spectacles results in poor detection and localization of ocular features. Separating the reflection components accurately is a relevant and important challenge in enhancing eye feature quality under spectacles. Direct application of Dichromatic reflection model (DRM) based reflection removal algorithms on real-time facial images under spectacles—results in loss of ocular feature information in the diffuse component—as the DRM approaches fails to separate the chromatic and achromatic part of diffuse reflection. In this manuscript, we solve the problem of hue-saturation ambiguity by extending the DRM to Hue-Saturation-Value (HSV) colour space, and propose a new single image-based spectacle problem removal approach. Experiments on three set of databases demonstrate that the proposed approach achieves better solution with minimum execution time compared to the state-of-the-art.