VAMamba：用于图像复原的高效视觉自适应 Mamba
VAMamba: An Efficient Visual Adaptive Mamba for Image Restoration
AI总结：
本文提出 VAMamba 框架，通过 QCLAM 动态缓存复用特征，并利用 GPS-SS2D 根据像素重要性自适应规划扫描路径，在多项图像复原任务上同时提升质量与效率。
摘要：
近期基于 Mamba 的图像复原方法已取得可喜成果，但仍受限于固定扫描模式和低效的特征利用。传统 Mamba 架构依赖预先确定的路径，无法适应多样化退化，从而限制复原性能和计算效率。为克服这些局限，我们提出 VAMamba，这是一个具有两项关键创新的视觉自适应 Mamba 框架。首先，QCLAM（Queue-based Cache Low-rank Adaptive Memory，基于队列的缓存低秩自适应记忆）通过存储历史表示的 FIFO 缓存增强特征学习。当前经 LoRA 适配的特征与缓存特征之间的相似性引导智能融合，实现动态复用，同时有效控制内存增长。其次，GPS-SS2D（Greedy Path Scan SS2D，贪心路径扫描 SS2D）引入自适应扫描。Vision Transformer 生成分数图以估计像素重要性，贪心策略确定最佳前向和后向扫描路径。这些学习到的轨迹取代僵化模式，使 SS2D 能够执行有针对性的特征提取。QCLAM 与 GPS-SS2D 的结合使 VAMamba 能够在保持高计算效率的同时，自适应地关注退化区域。在多种复原任务上的大量实验表明，VAMamba 在复原质量和效率方面始终优于现有方法，为自适应图像复原建立了新基准。我们的代码可在 https://github.com/WaterHQH/VAMamba 获取。
英文摘要：
Recent Mamba-based image restoration methods have achieved promising results but remain limited by fixed scanning patterns and inefficient feature utilization. Conventional Mamba architectures rely on predetermined paths that cannot adapt to diverse degradations, constraining both restoration performance and computational efficiency. To overcome these limitations, we propose VAMamba, a Visual Adaptive Mamba framework with two key innovations. First, QCLAM(Queue-basedCacheLow-rankAdaptiveMemory)enhancesfeaturelearningthrougha FIFO cache that stores historical representations. Similarity between current LoRA-adapted and cached features guides intelligent fusion, enabling dynamic reuse while effectively controlling memorygrowth.Second, GPS-SS2D(GreedyPathScanSS2D)introducesadaptive scanning. A Vision Transformer generates score maps to estimate pixel importance, and a greedy strategy de termines optimal forward and backward scanning paths. These learned trajectories replace rigid patterns, enabling SS2D to perform targeted feature extraction. The integration of QCLAM and GPS-SS2D allows VAMamba to adaptively focus on degraded regions while maintaining high computational efficiency. Extensive experiments across diverse restoration tasks demonstrate that VAMamba consistently outperforms existing approaches in both restoration quality and efficiency, establishing new benchmarks for adaptive image restoration. Our code is available at https://github.com/WaterHQH/VAMamba.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
