.png){kind=icon}
.png){kind=icon}
Return
Graph information bottleneck for remote sensing segmentation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2025.131662.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Remote sensing segmentation has a wide range of applications in environmental protection, urban change detection, etc. Despite the success of deep learning-based remote sensing segmentation methods (e.g., CNN and Transformer), they are not flexible enough to model irregular objects. In addition, existing graph contrastive learning methods usually adopt the approach of maximizing mutual information to keep the node representations consistent between different graph views, which may cause the model to learn task-independent redundant information (i.e., information unrelated to the downstream task, including both redundancy and noise.). To tackle the above problems, this paper treats images as graph structures and introduces a novel Graph Information Bottleneck for Remote Sensing Segmentation (GIB-RSS) architecture. Specifically, we construct a node-masking and edge-masking graph view to obtain an optimal graph structure representation, which can adaptively learn whether to mask nodes and edges. Here, the optimal graph structure representation refers to the refined node and edge embeddings derived from the masked graph views under the GIB objective, where task-relevant structural information is preserved while task-irrelevant redundancy and noise are suppressed. Furthermore, this paper innovatively introduces information bottleneck theory into graph contrastive learning to maximize task-related information while minimizing task-independent redundant information. Finally, we replace the convolutional module in UNet with the GIB-RSS module to complete the segmentation and classification tasks of remote sensing images. Extensive experiments on publicly available real datasets demonstrate that our method outperforms state-of-the-art remote sensing image segmentation methods.
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
View more organizations