EVENT DOSSIER
On September 7, 2026, arXiv cs.CV published a research paper titled “Fractional-Order Adaptive Motion Magnification: Phase-Reliability Weighting for Noise-Constrained Video Amplification”. This study proposes a fractional-order adaptive motion enhancement method aimed at solving the problem of video amplification under noise constraints. The core innovation of this method lies in the introduction of a phase reliability weighting mechanism, which is used to optimize the process of extracting and amplifying motion details in videos, thereby effectively suppressing noise interference while maintaining image quality.
研究人员提出 FrAM（Fractional-order Adaptive Motion Magnification）技术，旨在解决传统 Eulerian video amplification 在纹理缺失区域因相位不可靠而放大传感器噪声的问题。该技术将恒定时间增益替换为分数阶 Gr\"unwald--Letnikov 导数，并将均匀空间增益替换为基于单生成信号局部幅度的像素权重。在包含纹理和平坦区域的合成序列测试中，FrAM 保持了与基准相同的放大效果，同时将平坦区域的时间噪声降至输入水平，且该降噪效果跨越八倍输入噪声范围依然有效。实际视频显示 FrAM 提升了空间选择性并降低了背景噪声。此外，因果重构使每帧计算成本降低两个数量级，在 640$\times$480 分辨率下达到 69 帧每秒的处理速度。