[论文] PrismRestore: A Compact Restoration Toolbox for Agentic Image Restorat...
论文概要
研究领域: CV
作者: Zhaocheng Yu, Kui Jiang, Jing Cao
发布时间: 2025-11-15
arXiv: 2511.13008
中文摘要
Agent化图像恢复通过组合操作来处理混合退化,但其工具箱通常需要存储多个恢复网络。我们提出PrismRestore,从一个冻结的多功能一体化模型中构建一个紧凑的可调用恢复动作库。我们的方法建立在共享模型中任务相关但相互重叠的通道响应之上。离线任务先验搜索利用激活和梯度统计信息将具有相似响应特征的通道分组,然后在非代表性通道上搜索任务特定的调制系数,同时保持代表性通道不被调制。所得的先验库实现了显式的动作选择和组合,无需额外的恢复骨干网络或任务特定适配器模块。一个状态条件策略使用组相对策略优化(GRPO)来学习组合这些动作。其目标函数在同一输入采样的轨迹组内分别归一化保真度和感知反馈,并将终端评估与来自逐步质量变化的return-to-go相结合。在合成复合退化和真实世界多任务基准上的实验表明,PrismRestore优于所评估的一体化基线,恢复质量与多模型Agent具有竞争力。PrismRestore仅用单一恢复骨干网络就实现了这些结果,且推理速度比所评估的Agent化基线更快。
原文摘要
Agentic image restoration composes operations to handle mixed degradations, yet its toolbox often requires storing multiple restoration networks. We propose PrismRestore, which constructs a compact bank of callable restoration actions from a single frozen all-in-one model. Our approach builds on task-dependent yet overlapping channel responses in the shared model. An offline task-prior search uses activation and gradient statistics to group channels with similar response profiles, then searches task-specific modulation coefficients over non-representative channels while leaving representative channels unmodulated. The resulting prior bank enables explicit action selection and composition without additional restoration backbones or task特定 adapter modules. A state-conditioned policy learns t...
自动采集于 2026-11-20
#论文 #arXiv #CV #小凯