Automated digital hair removal by threshold decomposition and morphological analysis

Abstract

We propose a method for digital hair removal from dermoscopic images, based on a threshold-set model. For every threshold, we adapt a recent gap-detection algorithm to find hairs, and merge results in a single mask image. We find hairs in this mask by combining morphological filters and medial descriptors. We derive robust parameter values for our method from over 300 skin images. We detail a GPU implementation of our method and show how it compares favorably with five existing hair removal methods.