Accurate segmentation of wound images is critical for effective diagnosis, treatment planning, and monitoring of wound healing. In this study, we introduce an advanced deep learning framework for the simultaneous segmentation of wounds and scale markers. Our approach leverages state-of-the-art architectures—including U-Net++, DeepLabV3+, and Segformer—combined with pre-trained encoders and diverse transfer learning strategies (full fine-tuning, decoder-only tuning, and partial tuning) to address the challenges posed by limited and heterogeneous clinical datasets. Evaluations conducted on the NBC2025 Challenge dataset demonstrate the effectiveness of our method, with models achieving high F1-scores, including 0.8934 on the test dataset, and competitive mean Intersection over Union (mIoU) values on the validation set. Extensive data augmentation further enhances model generalization, ensuring robust segmentation performance. The promising results underscore the potential of deep learning-based segmentation as a reliable tool for objective wound assessment, thereby supporting improved clinical decision-making and patient care.