This article presents an integrated approach for automated detection and multi-class classification of dermoscopic skin lesions using digital image processing and artificial intelligence techniques. Skin cancer, particularly melanoma, remains one of the most aggressive malignancies, where early diagnosis significantly improves survival rates. The proposed framework incorporates artifact removal, color normalization, U-Net-based lesion segmentation, and transfer learning with pretrained convolutional neural networks. Experimental results demonstrate high segmentation accuracy (Dice = 0.89) and strong classification performance (ROC-AUC = 0.96). Statistical validation through cross-validation and confidence interval analysis confirms the robustness and generalization capability of the model.