SATB-VR: Training Few-Step Video Restoration Diffusion Model using SNR-Aware Trajectory Blending
This work addresses the efficiency-quality trade-off in video restoration diffusion models, offering a practical few-step solution for practitioners needing fast yet high-quality restoration.
SATB-VR trains a few-step video restoration diffusion model by using an auxiliary predictor to skip early low-SNR steps and a SNR-Aware Trajectory Blending strategy to resolve train-inference discrepancy, achieving favorable performance against existing methods with ≤5 steps on synthetic, real-world, and AIGC benchmarks.
While diffusion models excel in video restoration, their reliance on extensive iterative steps limits efficiency. Conversely, aggressive single-step distillation often compromises fine texture recovery. To achieve an optimal balance, we present SATB-VR, a few-step paradigm that jump-starts the denoising process via an auxiliary predictor, explicitly bypassing early low signal-to-noise ratio (SNR) steps. However, naive joint training of the predictor and the denoiser inherently introduces a severe train-inference discrepancy. To resolve this, we propose the SNR-Aware Trajectory Blending (SATB) strategy. During the forward process, SATB constructs the noisy input by dynamically blending the predictor's output with the ground-truth trajectory based on the SNRs. This forces the denoiser to robustly compensate for initial prediction errors while smoothly converging to the clean data manifold. Furthermore, we introduce a Denoiser-Driven Consistency (DDC) loss, leveraging the concurrently updated denoiser as a dynamic evaluator to explicitly align internal features and boost predictor accuracy. Extensive experiments demonstrate that, under flexible few-step inference regimes (\eg, $\le 5$ steps), SATB-VR performs favorably against existing approaches on synthetic, real-world, and AIGC benchmarks.