YODA: Yet Another One-step Diffusion-based Video Compressor

TL;DR

YODA introduces a novel one-step diffusion-based video compression method that leverages multiscale temporal features and a linear Diffusion Transformer, achieving state-of-the-art perceptual quality and outperforming existing methods.

Contribution

The paper proposes YODA, a new one-step diffusion model for video compression that incorporates temporal features and a linear Diffusion Transformer for improved spatial-temporal correlation exploitation.

Findings

YODA outperforms traditional and deep-learning baselines on perceptual metrics.

YODA achieves state-of-the-art results on LPIPS, DISTS, FID, and KID.

The source code will be publicly available.

Abstract

While one-step diffusion models have recently excelled in perceptual image compression, their application to video remains limited. Prior efforts typically rely on pretrained 2D autoencoders that generate per-frame latent representations independently, thereby neglecting temporal dependencies. We present YODA--Yet Another One-step Diffusion-based Video Compressor--which embeds multiscale features from temporal references for both latent generation and latent coding to better exploit spatial-temporal correlations for more compact representation, and employs a linear Diffusion Transformer (DiT) for efficient one-step denoising. YODA achieves state-of-the-art perceptual performance, consistently outperforming traditional and deep-learning baselines on LPIPS, DISTS, FID, and KID. Source code will be publicly available at https://github.com/NJUVISION/YODA.