Automatic Reflection Detection on Dermatological Images Acquired via Mobile Devices

The late detection of skin cancer, especially melanoma, leads to an increase in mortality rates, therefore it is high time to embrace mobile teledermatology and adapt the existing methodologies to check out the malignancy level of a mole. Though, the usage of a smartphone camera gives rise to new pre-processing challenges like the treatment of reflections caused by the flash light or uneven illumination conditions. In this work, a new methodology to detect reflections on mobile-acquired dermatological images is presented, based on the different behaviour in terms of highlights invariance of the L and H colour channels, from the CIE L*a*b* and HSV color spaces, respectively. The achieved results confirm the good quality of the suggested algorithm.