Segmentation of Vitiligo images using modified Watershed Segmentation

Abstract

Vitiligo is a dermatological condition characterized by depigmentation of the skin, resulting from the loss of melanocytes. Automatic image analysis methods are essential for diagnosing vitiligo and tracking its development. Because it is good at defining object boundaries, watershed segmentation is a frequently used technique for segmenting medical images. However, traditional watershed segmentation may produce over-segmentation and under-segmentation issues in vitiligo images due to their complex and irregular patterns.

This research paper proposes a modified watershed segmentation approach tailored specifically for vitiligo image analysis. The proposed method incorporates pre-processing steps to enhance image contrast and reduce noise, followed by a novel marker-based watershed segmentation algorithm. In this algorithm, markers are strategically placed based on local intensity and texture features to guide the segmentation process and improve boundary delineation.

Annotated ground truth images from a collection of vitiligo images are used to assess the effectiveness of the suggested technique. Several quantitative indicators, including accuracy, the Jaccard index, and the Dice similarity coefficient, are used to compare the segmentation outcomes produced by the suggested method with those produced by advanced techniques.

Experimental results demonstrate that the modified watershed segmentation method outperforms existing approaches in accurately segmenting vitiligo lesions while minimizing over-segmentation and under-segmentation errors. The proposed method shows promising potential for aiding dermatologists in diagnosing and monitoring vitiligo progression, facilitating timely treatment interventions, and improving patient confidence.