Related Experiment Video
Updated: May 19, 2026
05:39
Dermoscopy Aids in the Diagnosis of Discoid Lupus Erythematosus
Published on: May 16, 2025
A perceptually oriented method for contrast enhancement and segmentation of dermoscopy images
Qaisar Abbas1, Irene Fondón Garcia, M Emre Celebi
1Department of Computer Science, National Textile University, Faisalabad, Pakistan. drqaisar@ntu.edu.pk
Summary
This study presents a novel method for automatic melanoma border detection (MBD) in dermoscopy images, significantly improving accuracy for lesion recognition. The approach enhances image contrast and uses advanced segmentation for precise border identification.
Area of Science:
- Dermatology
- Medical Imaging
- Computer Vision
Background:
- Dermoscopy images often have low contrast due to lighting variations, hindering accurate lesion border detection.
- Automatic melanoma border detection (MBD) is critical for reliable lesion recognition.
Purpose of the Study:
- To develop a novel, perceptually oriented approach for accurate melanoma border detection (MBD).
- To enhance lesion contrast and segmentation in dermoscopy images for improved MBD.
Main Methods:
- A four-step MBD system combining region and edge-based segmentation.
- Transformation to CIE L*a*b* color space for contrast enhancement.
- Hill-climbing algorithm for region-of-interest detection and adaptive thresholding for border determination.
Main Results:
- The proposed MBD method achieved a true positive rate (TPR) of 94.25%, false positive rate (FPR) of 3.56%, and error probability (EP) of 4% on 100 dermoscopy images.
- Outperformed three state-of-the-art segmentation techniques in comparative analysis.
Conclusions:
- The developed MBD approach demonstrates high accuracy in detecting lesion borders.
- Software and sample images are available for download, facilitating further research and application.
