The core message of this paper is that leveraging fine-grained information and decoupling task-specific and task-agnostic noise are crucial for accurate remote sensing change detection. The authors propose a series of operations called FINO (Fine-grained Information compensation and Noise decOupling) to address these challenges.
A novel hybrid change encoder that leverages both local and global feature representations to precisely detect subtle and large change regions in bi-temporal remote sensing images.
MFDS-Net proposes a multi-scale feature depth-supervised network that enhances the processing of global semantic information and local detail features to achieve robust and accurate change detection in remote sensing imagery.
This paper surveys the recent advancements in applying foundation models, particularly those pre-trained on large datasets and fine-tuned for specific tasks, to the challenge of change detection in remote sensing imagery.