标题： 拉普拉斯-马巴：拉普拉斯频率先验引导的马巴-CNN融合网络用于图像去雾 显示英文标题

标题： Laplace-Mamba: Laplace Frequency Prior-Guided Mamba-CNN Fusion Network for Image Dehazing

摘要： 最近在图像修复方面的进展强调了空间状态模型（SSMs）作为建模长距离依赖关系的强大工具，由于其吸引人的线性复杂度和计算效率。然而，基于SSMs的方法在重建局部结构方面存在局限性，并且在处理高维数据时效果较差，经常导致细粒度图像特征的次优恢复。为了解决这些挑战，我们引入了Laplace-Mamba，这是一种新的框架，将拉普拉斯频率先验与混合Mamba-CNN架构结合，以实现高效的图像去雾。利用拉普拉斯分解，图像被解耦为捕捉全局纹理的低频成分和表示边缘和细粒度细节的高频成分。这种分解通过双并行路径实现了专门处理：低频分支使用SSMs进行全局上下文建模，而高频分支使用CNN来细化局部结构细节，有效应对各种雾霾场景。值得注意的是，拉普拉斯变换根据奈奎斯特理论实现了低频成分的信息保留下采样，从而显著提高了计算效率。在多个基准上的广泛评估表明，我们的方法在恢复质量和效率方面都优于最先进的方法。源代码和预训练模型可在https://github.com/yz-wang/Laplace-Mamba获取。 显示英文摘要

摘要： Recent progress in image restoration has underscored Spatial State Models (SSMs) as powerful tools for modeling long-range dependencies, owing to their appealing linear complexity and computational efficiency. However, SSM-based approaches exhibit limitations in reconstructing localized structures and tend to be less effective when handling high-dimensional data, frequently resulting in suboptimal recovery of fine image features. To tackle these challenges, we introduce Laplace-Mamba, a novel framework that integrates Laplace frequency prior with a hybrid Mamba-CNN architecture for efficient image dehazing. Leveraging the Laplace decomposition, the image is disentangled into low-frequency components capturing global texture and high-frequency components representing edges and fine details. This decomposition enables specialized processing via dual parallel pathways: the low-frequency branch employs SSMs for global context modeling, while the high-frequency branch utilizes CNNs to refine local structural details, effectively addressing diverse haze scenarios. Notably, the Laplace transformation facilitates information-preserving downsampling of low-frequency components in accordance with the Nyquist theory, thereby significantly improving computational efficiency. Extensive evaluations across multiple benchmarks demonstrate that our method outperforms state-of-the-art approaches in both restoration quality and efficiency. The source code and pretrained models are available at https://github.com/yz-wang/Laplace-Mamba.