Effective Border Detection of Noisy Real Skin Lesions for Skin Lesion Diagnosis by Robust Segmentation Algorithm

Abstract

One of the most important steps in image analysis is the automated detection of lesion borders. Early detection of melanoma is one of
the greatest challenges of dermatological practice today. Accurate image segmentation of skin lesions is one of the key steps for useful, early,
and non-invasive diagnosis of coetaneous melanomas. The medical images generally are bound to contain noise while acquisition. This paper
proposes a robust and efficient image segmentation algorithm to extract the true border of noisy clinical skin images containing lesions, that
reveals the global structure irregularity, which may suggest excessive cell growth or regression of a melanoma. The proposed segmentation
algorithm is applied to a RGB image containing the lesion, where the RGB image is converted to grayscale intensity image by eliminating the
hue and saturation information while retaining the luminance and adds salt and pepper noise to the image and uses background noise reduction
techniques to filter noise. The proposed algorithm converts an image to a binary image, based on threshold, and finds edges in the image using
canny method and traces the object in the image. To verify the capability of the segmentation algorithm in detecting the borders of the lesions for
skin lesion diagnosis, the algorithms was applied on diversity of clinical skin images containing lesions. The results demonstrate the successful
border detection of real skin lesions by the proposed segmentation algorithm for noisy clinical skin images and make them accessible for further
analysis and research.

 

Keywords: Image Segmentation, Skin Lesion, Canny detector, Border detection, Melenoma