arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2611.04416cs.CV
收藏

基于稀疏跨分辨率去噪的高效且保真高分辨率图像编辑

Efficient and Faithful High-Resolution Image Editing with Sparse Cross-Resolution Denoising

Shanghai Innovation Institute(上海创新研究院);Fudan University(复旦大学);StepFun(阶跃星辰);The Chinese University of Hong Kong(香港中文大学)

Maosen Zhao, Xianfang Zeng, Pengtao Chen, Peng Ye, Gang Yu, Jiayuan Fan, Tao Chen

AI总结:

提出FLASH编辑器,通过稀疏跨分辨率去噪实现高效高保真2K/4K图像编辑,在VINS-4KEval上结构保真度最优,推理速度显著提升且延迟降低。

摘要:

基于指令的图像编辑器目前主要在1K分辨率下运行,而实际编辑场景越来越需要保真的2K/4K输出。密集高分辨率适配能保留编辑轨迹,但会产生高昂的目标-源注意力开销;先在1K分辨率编辑再进行外部细化的方案成本更低,但会将细节恢复与编辑轨迹分离。我们提出FLASH,一种基于稀疏跨分辨率去噪的原生高分辨率编辑器。冻结的1K分支提供语义和结构锚点,而目标分辨率分支在同一Transformer内重建局部细节。非对称拓扑保留完整1K注意力,通过空间对齐的低分辨率噪声目标状态引导高分辨率查询,并将高分辨率目标-参考交互限制在局部Halo邻域内。特定分辨率的扩散状态映射和空间YaRN将预训练编辑器适配到更大的网格。可选的无训练渐进路径用1K前缀替换早期高分辨率去噪,并在同一编辑器中完成细化。在VINS-4KEval数据集上,直接模型在2K和4K分辨率下均达到对比方法中最强的结构保真度。单独的4K前向基准(R=1)显示,总前向执行速度比密集目标-参考注意力快4.2倍,注意力速度快25.3倍。渐进推理将端到端4K延迟降低30.5%,仅伴随适度的质量损失。

英文摘要:

Instruction-based image editors still primarily operate at 1K, whereas practical editing increasingly requires faithful 2K/4K outputs. Dense high-resolution adaptation preserves the editing trajectory but incurs expensive target-source attention; editing at 1K followed by external refinement is cheaper but can separate detail restoration from the editing trajectory. We introduce FLASH, a native high-resolution editor based on sparse cross-resolution denoising. A frozen 1K branch provides a semantic and structural anchor, while a target-resolution branch reconstructs local detail within the same Transformer. An asymmetric topology retains full 1K attention, guides high-resolution queries with spatially aligned low-resolution noisy-target states, and confines high-resolution target-reference interaction to local Halo neighborhoods. Resolution-specific diffusion-state mapping and spatial YaRN adapt the pretrained editor to larger grids. An optional, training-free progressive path replaces early high-resolution denoising with a 1K prefix and completes refinement in the same editor. On VINS-4KEval, the direct model achieves the strongest structural fidelity among the compared methods at both 2K and 4K. A separate 4K forward benchmark with R = 1 records 4.2x faster total forward execution and 25.3x faster attention than dense target-reference attention. Progressive inference reduces end-to-end 4K latency by 30.5%, with a modest quality cost.

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。

相关深度报道

↑