Automated Segmentation of Skin Lesion Images using Dominant Intensity Pixel: A Boundary Detection Technique

Abstract

Dermoscopy is one of the major imaging modalities used in the diagnosis of melanoma and other pigmented skin lesions. Automated segmentation of dermoscopy images is an important step for computer-aided diagnosis of melanoma. In this paper, we investigate how to use the dominant intensity pixel present in the border of the pigmented lesions to improve the segmentation of dermoscopy images. The algorithm groups the dominant intensity image pixels with homogeneous properties, and merges the pixel groups into a few super-regions. The optimal lesion skin boundary is chosen from the set of all region boundaries, where the optimality is determined from the color and texture properties of the regions. We test our method on many dermoscopy images and compare the automatically generated segmentation with dermatologist-determined segmentation. The results demonstrate the advantage of incorporating intensity specific constraints into the segmentation process. The medical images, while acquisition is generally bound to contain noise. The preprocessing stage employs background noise reduction techniques to filter noise. The algorithm converts an image to a binary image, based on threshold, and finds edges in the image using dominant intensity and traces the object boundary in the image. The algorithm was applied on many noisy clinical skin images containing lesions. The proposed segmentation algorithm successfully detects the border of noisy clinical skin images and makes them accessible for further analysis and research.

 

Keywords: Border detection, dermoscopy, Image Segmentation, Skin Lesion