CEBSNet: Change-Excited and Background-Suppressed Network with Temporal Dependency Modeling for Bitemporal Change Detection

TL;DR

CEBSNet tackles bitemporal change detection by explicitly modeling temporal dependencies and balancing emphasis between prominent and subtle changes. It introduces a Channel Swap Module (CSM) to capture cross-temporal correlations, a dual-branch Feature Excitation and Suppression Module (FESM) to excite change regions while suppressing background, and a Pyramid-Aware Spatial-Channel Attention (PASCA) to capture multi-scale changes. Through Change-Guided Feature Fusion (CGFF) and a multi-scale decoder, CEBSNet delivers robust change maps across street-view and remote-sensing datasets, achieving state-of-the-art performance with competitive computational cost. The work demonstrates that temporal interaction and targeted feature refinement significantly enhance both detection accuracy and localization, with potential for efficient deployment in real-world monitoring tasks.

Abstract

Change detection, a critical task in remote sensing and computer vision, aims to identify pixel-level differences between image pairs captured at the same geographic area but different times. It faces numerous challenges such as illumination variation, seasonal changes, background interference, and shooting angles, especially with a large time gap between images. While current methods have advanced, they often overlook temporal dependencies and overemphasize prominent changes while ignoring subtle but equally important changes. To address these limitations, we introduce \textbf{CEBSNet}, a novel change-excited and background-suppressed network with temporal dependency modeling for change detection. During the feature extraction, we utilize a simple Channel Swap Module (CSM) to model temporal dependency, reducing differences and noise. The Feature Excitation and Suppression Module (FESM) is developed to capture both obvious and subtle changes, maintaining the integrity of change regions. Additionally, we design a Pyramid-Aware Spatial-Channel Attention module (PASCA) to enhance the ability to detect change regions at different sizes and focus on critical regions. We conduct extensive experiments on three common street view datasets and two remote sensing datasets, and our method achieves the state-of-the-art performance.

Figures (10)

• Figure 1: Motivation analysis. We compare our CEBSNet with existing CD methods on VL-CMU-CD and LEVIR-CD datasets. The objects in red dotted box are important change regions, and the F1 scores objectively measure the performance.
• Figure 2: The architecture of our CEBSNet. It comprises three key stages: 1) Spatial-Temporal Feature Extraction, where a siamese encoder with Channel Swap Module (CSM) extracts multi-scale spatiotemporal feature pairs; 2) Change Feature Refinement, where change features are generated and iteratively refined through CGFF, FESM and PASCA; and 3) Multi-Scale Change Detection, where refined change features are decoded into multi-scale change maps and fused to generate the final change map.
• Figure 3: The architecture of CSM. It exchanges channel-level features between image pairs to model temporal dependencies.
• Figure 4: The architecture of CGFF. It adaptively fuses bi-temporal features guided by refined change features through bi-temporal concatenation, absolute difference computation, and change-guided feature fusion.
• Figure 5: The architecture of FESM. It consists of two branches: the excitation branch enhances prominent change regions, while the suppression branch captures subtle changes by boosting foreground and suppressing background.