Treffer: Saliency feature fusion constraint for skin lesion level set segmentation.

Skin lesion segmentation is challenging due to artifacts, hair, low contrast between lesions and their boundaries, color noise, and other interferences. To address these issues, we propose a novel saliency feature fusion constraint level set segmentation model (SFFC-LSM). First, a skin lesion detection model based on saliency feature fusion is developed. This model utilizes colour and contrast features for the initial segmentation of the lesion region, providing prior shape information. This prior information is then incorporated as a constraint term into the fine segmentation level set method, and the level set energy function is improved. Finally, precise skin lesion segmentation is achieved by minimizing the energy function. Experimental results show that, when tested on the ISIC2016 and ISIC2018 datasets, the proposed method effectively overcomes interference and outperforms other advanced segmentation algorithms, achieving accurate skin lesion segmentation. [ABSTRACT FROM AUTHOR]