Hair removal on dermoscopy images

Abstract

Digital Dermoscopy is a tool commonly used by dermatologists for assisting the diagnosis of skin lesions. The presence of hair in such dermoscopic images frequently occludes significant diagnostic information and reduces their value. In this work we propose algorithms that successfully identify and remove hair from the dermoscopic images. The proposed algorithms consist of two parts; the first deals with the identification of hair, while the second part concerns the image restoration using interpolation. For the evaluation of the algorithms we used ground truth images with synthetic hair and compared the results with the commonly used in the literature DullRazor tool. According to the experimental results the proposed hair removal algorithms can be used successfully in the detection and removal of both dark and light colored hair. © 2015 IEEE.