To fully utilize spatial information for segmentation and address the challenge of handling areas with significant grayscale variations in remote sensing segmentation, we propose the spatial and frequency domain fusion network (SFFNet) framework. This framework employs a two-stage network design: the first stage extracts features using spatial methods to obtain features with sufficient spatial details and semantic information; the second stage maps these features in both spatial and frequency domains. In the frequency domain mapping, we introduce the wavelet transform feature decomposer (WTFD) structure, which decomposes features into low-frequency and high-frequency components using the Haar wavelet transform and integrates them with spatial features. To bridge the semantic gap between frequency and spatial features, facilitating significant feature selection to promote the combination of features from different representation domains, we design the multiscale dual-representation alignment filter (MDAF). This structure utilizes multiscale convolutions and dual-cross attentions. Comprehensive experimental results demonstrate that, compared to existing methods, SFFNet achieves superior performance in terms of mean intersection over union (mIoU), reaching 84.80% and 87.73%, respectively. The code is located at https://github.com/yysdck/SFFNet .

In the field of remote sensing change detection, accurately capturing temporal change information and efficiently integrating multi-level information is a major challenge. In order to extend the sensory domain and optimize the information fusion, the model is able to capture Temporal-Spatial change features more accurately and improve the accuracy of change detection. In this paper, we propose a Temporal-Spatial Multi-scale Graph Attention Network (TSMGA), specifically, TSMGA employs a pair of pre-trained ResNet18 for effective multiscale feature extraction, and in order to enhance the disparity information of the bi-temporal images, we also design the Temporal Fusion Block (TFB) to emphasize the changed areas. The spatial and channel features of multiscale disparity features are enhanced by Multiscale Spatial-Channel Aggregation Module(MSCAM). To enable more robust and efficient exploration of more global contextual information, we are inspired to introduce Shortest Path Graph Attention (SPGA), which allows for a more informative and complete exploration of the global context, and furthermore allows for more efficient gathering information from far-off neighbors to the central node. In order to ensure comprehensive utilization of local and global features and to significantly improve the clarity and detail retention of the output image, Global Context Residual Fusion Module (GCRFM) is designed, GCRFM efficiently utilizes the complementary information of the feature maps for fusing and recovering spatial details and variation information. We validate the effectiveness and advancement of TSMGA on three classical datasets (LEVIR-CD, WHU-CD, GZ-CD), and the experimental results show that TSMGA achieves the state-of-the-art performance level.

Accurate remote sensing change detection (RSCD) tasks rely on comprehensively processing multi-scale information from local details to effectively integrate global dependencies. Hybrid models based on convolutional neural networks (CNNs) and Transformers have become mainstream approaches in RSCD due to their complementary advantages in local feature extraction and long-term dependency modeling. However, the Transformer faces application bottlenecks due to the high secondary complexity of its attention mechanism. In recent years, state-space models (SSMs) with efficient hardware-aware design, represented by Mamba, have gained widespread attention for their excellent performance in long-series modeling and have demonstrated significant advantages in terms of improved accuracy, reduced memory consumption, and reduced computational cost. Based on the high match between the efficiency of SSM in long sequence data processing and the requirements of the RSCD task, this study explores the potential of its application in the RSCD task. However, relying on SSM alone is insufficient in recognizing fine-grained features in remote sensing images. To this end, we propose a novel hybrid architecture, ConMamba, which constructs a high-performance hybrid encoder (CS-Hybridizer) by realizing the deep integration of CNN and SSM through the feature interaction module (FIM). In addition, we introduce the Spatial Integration Module (SIM) in the feature reconstruction stage to further enhance the model’s ability to integrate complex contextual information. Extensive experimental results on three publicly available RSCD datasets show that ConMamba significantly outperforms existing techniques in several performance metrics, validating the effectiveness and foresight of the hybrid architecture based on CNN and SSM in remote sensing change detection.

Automatic crack detection is challenging due to the poor continuity of cracks, the different widths of cracks, and the low contrast between cracks and the surrounding pavement. In this paper, a deep convolutional neural network called CurSeg is proposed, which achieves pixelwise segmentation of cracks in an end‐to‐end manner. In this approach, features at different scales are fused together to attain the context information from the cracks. The elaborately designed model can effectively suppress the propagation of noise and further refine the crack features by aggregating multiscale and multilevel features from low‐level to high‐level. Residual detail attention (RDA) is also introduced to better capture the line structure and the ability to accurately locate the crack position in a complex context to make the network more discriminative and robust. CurSeg is evaluated on four datasets to validate the effectiveness of the approach. The experimental results demonstrate that this method achieves state‐of‐the‐art performance on the four challenging datasets.

Multifocus image fusion is the merging of images of the same scene and having multiple different foci into one all-focus image. Most existing fusion algorithms extract high-frequency information by designing local filters and then adopt different fusion rules to obtain the fused images. In this paper, a wavelet is used for multiscale decomposition of the source and fusion images to obtain high-frequency and low-frequency images. To obtain clearer and complete fusion images, this paper uses a deep convolutional neural network to learn the direct mapping between the high-frequency and low-frequency images of the source and fusion images. In this paper, high-frequency and low-frequency images are used to train two convolutional networks to encode the high-frequency and low-frequency images of the source and fusion images. The experimental results show that the method proposed in this paper can obtain a satisfactory fusion image, which is superior to that obtained by some advanced image fusion algorithms in terms of both visual and objective evaluations.