Multiscale Morphological Reconstruction for Hair Removal in Dermoscopy Images
November 2018
in “
Modern Applied Science
”
TLDR The method accurately detects and removes hair from skin images to improve melanoma diagnosis.
The paper "Multiscale Morphological Reconstruction for Hair Removal in Dermoscopy Images" from 2018 proposed a method for detecting and removing hair from dermoscopy images, which can distort the diagnosis of melanoma. The method used a convolution of the image with a kernel from the first derivative of the Gaussian function and replaced the hairs using a multiscale morphological reconstruction. A refining stage was also integrated to maintain the quality of the patterns on the lesion. The method was tested on 36 dermoscopy images, which included a total of 586 hairs. The results showed sensitivity and specificity performance measurements of 94.14% and 99.89%, respectively, indicating a high level of accuracy in detecting and removing hair from the images.
