Automatic skin lesion segmentation with optimal colour channel from dermoscopic images

Abstract

This article provides an improved automated skin lesion segmentation method for dermoscopic images. There are several stages for this method. These include the pre-processing steps such as resizing the images and eliminating noise. Hair was removed and reflective light was reduced using morphological operations and a median filter. The single green channel was rescaled into new intensities, as it provided the highest segmentation accuracy. The threshold value was calculated to separate the skin lesion region from healthy skin. Morphological operations were implemented to merge the small lesion areas around the bigger lesion areas with similar features and trace the boundary of the melanoma. The accuracy of the segmentation was evaluated by comparing the automatic boundary and manual boundary. Compared to other studies, our proposed method achieved the highest average accuracy of 97%.