Abstract
Video restoration aims to reconstruct high quality video sequences from low quality inputs, addressing tasks such as super resolution, denoising, and deblurring. Traditional regression based methods often produce unrealistic details and require extensive paired datasets, while recent generative diffusion models face challenges in ensuring temporal consistency. We introduce DiTVR, a zero shot video restoration framework that couples a diffusion transformer with trajectory aware attention and a wavelet guided, flow consistent sampler. Unlike prior 3D convolutional or frame wise diffusion approaches, our attention mechanism aligns tokens along optical flow trajectories, with particular emphasis on vital layers that exhibit the highest sensitivity to temporal dynamics. A spatiotemporal neighbour cache dynamically selects relevant tokens based on motion correspondences across frames. The flow guided sampler injects data consistency only into low-frequency bands, preserving high frequency priors while accelerating convergence. DiTVR establishes a new zero shot state of the art on video restoration benchmarks, demonstrating superior temporal consistency and detail preservation while remaining robust to flow noise and occlusions.
Abstract (translated)
视频恢复的目标是从低质量的输入中重建高质量的视频序列，解决超分辨率、去噪和去模糊等问题。传统的回归方法通常会产生不真实的细节，并且需要大量的配对数据集，而最近的生成扩散模型在确保时间一致性方面面临挑战。我们引入了DiTVR（Diffusion Transformer with Trajectory-aware Video Restoration），这是一种零样本视频恢复框架，结合了一个具有轨迹感知注意力机制和小波引导、流一致采样的扩散变换器。 与之前的3D卷积或帧级扩散方法不同，我们的注意机制沿光学流动轨迹对齐令牌，并特别强调最敏感于时间动态的关键层。一个时空邻域缓存根据帧之间的运动对应关系动态选择相关令牌。受流引导的采样器仅在低频带中注入数据一致性，同时保留高频先验并加速收敛。 DiTVR在视频恢复基准测试上建立了新的零样本状态最佳表现，展示了卓越的时间一致性和细节保持能力，并且对流动噪声和遮挡具有鲁棒性。
URL
https://arxiv.org/abs/2508.07811
