Skin lesion segmentation plays a major role in medical image analysis and is important in identifying and segmenting skin lesions from surrounding tissue. Its accuracy is paramount for early detection and diagnosis of skin cancer and quantitative analysis of skin lesions. Although considerable progress has been made through the advancement of the deep learning models, including LinkNet and U-Net, challenges still exist for necessitating further improvements in several key areas. The challenges in this include handling different types of skins, the requirement to be effective while light conditions change, and making fewer false negatives. All these are very important to ensuring the reliability and robustness of the skin lesion segmentation task. In this research work, we propose a novel approach to solving the problem based on ensemble techniques that tap into their strength. We thoroughly train and test an extensive selection of deep learning models on the HAM100000 dataset that boasts a rich collection of skin lesion images. Our results show the effectiveness of our approach proposed above and are superior to the metrics of IOU. Additionally, it also mitigates the inherent flaws associated with standalone models.