<![CDATA[In the field of medical image analysis, artifacts such as dermal hair pose a major challenge to both visual interpretation and automated image processing during dermoscopic examinations. Hair covering the lesion area can obscure the lesion boundaries, reduce the quality of feature extraction, and lead to segmentation and classification errors. Recent studies have shown that dermal hair remains one of the most persistent artifacts affecting automated analysis, even in state-of-the-art segmentation models. These artifacts also degrade the performance of AI-based systems that rely on visual information. This study aims to improve the accuracy of hair segmentation in dermoscopic images through the application of effective and efficient preprocessing techniques. This study applies Principal Component Analysis (PCA) as a grayscale method to reduce the computational burden while preserving essential image features, and Contrast-Limited Adaptive Histogram Equalization (CLAHE) to enhance local contrast and highlight thin or low-contrast hair structures. The combination of PCA and CLAHE serves as a preprocessing stage to improve the quality of input images for deep learning-based segmentation models. The main contribution of this research is the integration of PCA-based grayscale methods with CLAHE in a single preprocessing pipeline before deep learning segmentation and the evaluation of their effects on the performance of the segmentation model. The evaluation is conducted using the AttU-Net architecture with Dice Similarity Coefficient (DSC) and Jaccard Index (JAC) metrics. The proposed PCA–CLAHE preprocessing achieves DSC and JAC values of 75.24% and 61.04%, respectively, outperforming the model without preprocessing. These results indicate that PCA–CLAHE effectively improves image quality and segmentation accuracy while maintaining computational efficiency.]]>
