Answers are grounded in this paper and cite their sources.
Multi-focus image fusion is crucial in various fields, particularly in satellite images, because it creates intricate images with augmented clarity. However, maintaining the original structure is challenging, leading to a reduction in performance. Therefore, this research proposes a Spatial Attention ResNet-50 Network (SARA-Net) model for fusing multiple satellite images. The proposed model utilizes an integration of ResNet-50 and Spatial Attention to effectively extract spatially significant features from actual images. Initially, satellite images from the publicly available SpaceNet datasets are preprocessed by normalization and grayscale conversion to provide consistency. ResNet-50 handles multi-scale feature extraction by residual learning, which addresses vanishing gradient issues, whereas spatial attention assigns adaptive weights to improve relevant spatial information within the feature map. The fusion layer combines these attention features to generate high-resolution and detail-preserving images. On the ten held-out test images, the proposed SARA-Net achieves a PSNR of 48.972 dB, an SSIM of 0.9953, and an entropy value of 8.4319, which is better than the existing methods.
No takes yet. Share an insight, caveat, or question.
Surendara et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: