Optimization of Skin Disease Image Segmentation: A Hybrid Approach Using HE-LAB Color Space with Canny and Otsu Methods

Abstract

Skin lesion detection through image analysis requires accurate segmentation methods to distinguish lesion regions from surrounding healthy skin. This study evaluates a hybrid approach that combines contrast enhancement using Histogram Equalization (HE) and HE applied to the Lightness (L) channel in the LAB color space (HE+LAB) with two segmentation methods, namely Canny edge detection and Otsu thresholding. The segmentation performance was evaluated at three image resolutions: 64 × 64, 128 × 128, and 256 × 256 pixels, using Accuracy, Precision, Recall, F1-Score, and Intersection over Union (IoU). The experimental results show that the HE+LAB-Otsu combination consistently achieves higher performance than the other evaluated combinations across the three resolutions. At 128 × 128 pixels, HE+LAB-Otsu achieves an Accuracy of 0.7170, F1-Score of 0.7407, and IoU of 0.5882. Canny-based segmentation generally produces lower scores, particularly for Recall and IoU, indicating difficulties in extracting complete lesion regions. The use of HE in the LAB color space improves lesion contrast while preserving the chromatic components of the image, resulting in better segmentation performance than standard HE in the evaluated experiments. The results indicate that the combination of HE+LAB and Otsu thresholding is a promising conventional approach for skin lesion image segmentation. However, further evaluation using additional datasets, statistical testing, and comparisons with modern deep learning segmentation methods is required to assess its generalizability and applicability to automated dermatological image analysis.