Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a hybrid Mamba-Transformer (MT) framework for remote sensing image super-resolution, aiming to overcome the limitations of CNNs and transformers in capturing long-range dependencies and maintaining computational efficiency. MT combines a focused mamba block (FMB) with a snake vision state-space module (SVSSM) for global feature modeling and a pixel-adaptive block (PAB) for pixel-level multiscale enhancement. Experiments on benchmark datasets demonstrate that MT outperforms state-of-the-art methods, achieving a better trade-off between performance and computational cost, specifically reducing parameters and FLOPs compared to MambaIRv2 while improving PSNR.
Achieve state-of-the-art remote sensing image super-resolution with a Mamba-Transformer hybrid that's 65% smaller and 68% faster than existing Mamba-based approaches.
Deep learning methods, such as convolutional neural networks (CNNs) and transformers, have achieved remarkable progress in remote sensing image (RSI) super-resolution (SR). However, CNNs struggle to capture long-range dependencies effectively, while transformers suffer from high computational overhead due to their quadratic complexity. To address these limitations, we propose a hybrid Mamba–transformer (MT) framework that simultaneously enhances global modeling efficiency and local detail reconstruction. MT consists of two core components: a focused mamba block (FMB) for global feature modeling and a pixel-adaptive block (PAB) for pixel-level multiscale enhancement. Within FMB, we introduce the snake vision state-space module (SVSSM), equipped with a snake selective scan module (SSSM) to strengthen long-range dependency modeling while reducing redundant computations, significantly improving global modeling efficiency. Meanwhile, PAB adaptively aggregates multiscale spatial information at the pixel level to more effectively restore fine-grained textures. Extensive experiments on multiple benchmark datasets demonstrate that MT consistently outperforms the existing state-of-the-art (SOTA) methods. Notably, compared with the latest MambaIRv2, MT reduces parameters by 65.3%, decreases FLOPs by 68.6%, and achieves a 0.10-dB improvement in average PSNR, achieving a superior balance between performance and computational efficiency.