Video Colorization with Pre-trained Text-to-Image Diffusion Models

Anonymous Authors


Video colorization is a challenging task that involves inferring plausible and temporally consistent colors for grayscale frames. In this paper, we present ColorDiffuser, an adaptation of a pre-trained text-to-image latent diffusion model for video colorization. With the proposed adapter-based approach, we repropose the pre-trained text-to-image model to accept input grayscale video frames, with the optional text description, for video colorization. To enhance the temporal coherence and maintain the vividness of colorization across frames, we propose two novel techniques: the Color Propagation Attention and Alternated Sampling Strategy. Color Propagation Attention enables the model to refine its colorization decision based on a reference latent frame, while Alternated Sampling Strategy captures spatiotemporal dependencies by using the next and previous adjacent latent frames alternatively as reference during the generative diffusion sampling steps. This encourages bidirectional color information propagation between adjacent video frames, leading to improved color consistency across frames. We conduct extensive experiments on benchmark datasets, and the results demonstrate the effectiveness of our proposed framework. The evaluations show that ColorDiffuser achieves state-of-the-art performance in video colorization, surpassing existing methods in terms of color fidelity, temporal consistency, and visual quality.

Colorization Results


  • We extend the pre-trained Stable Diffusion model to a reference-based frame colorization model. With the adapter-based mechanism, we obtain a conditional text-to-image latent diffusion model that leverages the power of the pre-trained Stable Diffusion model to render colors in the latent space \(z_c\), according to the visual semantics of the grayscale input \(g=\mathcal{E_g}(I_g)\), the text input, and the reference color latent \(z_{\texttt{ref},t}\).
  • During the inference, for each frame in the input grayscale video, we perform a parallel sampling process. Each sampling step for a particular frame is conditioned on the latent information from the previous sampling step of an adjacent frame. Essentially, the Color Propagation Attention and the Alternated Sampling Strategy coordinate the reverse diffusion process and enable bidirectional propagation of color information between adjacent frames to ensure consistency in colorization over time.