A UV Spot Segmentation Method Based on Facial Skin Image

Facial UV spot segmentation is an important research topic in the field of dermatology and aesthetics, and its results can provide a scientific basis for skin health assessment, early diagnosis of diseases, and the development of personalized skin care programs. However, due to the significant diversity of UV spots in different individuals and the influence of factors such as light variations, complex skin texture and noise interference, there are still many challenges to realize high-precision automatic segmentation. In this study, a pair of paired datasets are constructed, which include facial UV images and corresponding binary masks. Based on the dataset, this paper proposes a deep learning network WD-UNet specialized for facial UV spot segmentation, which aims to overcome the above difficulties and improve the segmentation performance. The network is based on ID-UNet's deep densely connected network and designed with a Wavelet transform channel attention feature extraction module (WTCA) and a orthogonal efficient hybrid attention feature fusion module (OEHA), the above modules facilitate the fusion of deep semantic information and shallow edge information in the dense connectivity across layers. Experiments show that our method outperforms existing segmentation methods in the UV spot segmentation task.