基于深度学习的二分图像分割综述OA

Dichotomous image segmentation is a crucial task in the field of computer vision,aiming to accurately seg-ment target objects from natural images to provide foundational support for visual analysis and scene understanding.In recent years,deep learning-based dichotomous segmentation has advanced significantly with many algorithms.However,there remains a lack of comprehensive surveys that systematically analyze and summarize these deep learning-driven approaches.Firstly,a thorough review of existing deep learning-based dichotomous segmentation methods is elaborated and they are classified into three categories:based on global-local information,auxiliary information,and diffusion models.Subsequently,benchmark datasets and evaluation metrics in the field are elaborated,and systematic quantitative and quali-tative experimental comparisons are conducted on the above three types of representative algorithms based on these met-rics.Finally,the challenges in dichotomous image segmentation are summarized and an outlook on its future development directions is provided.