SSI-DM: Singularity Skipping Inversion of Diffusion Models
This provides a principled and efficient solution for image editing tasks using diffusion models, addressing a specific bottleneck in inversion accuracy.
The paper tackled the problem of inverting real images into noise space for editing with diffusion models, which existing methods handled poorly due to a mathematical singularity, and proposed SSI-DM to bypass this by adding small noise, achieving superior performance on public datasets for reconstruction and interpolation tasks.
Inverting real images into the noise space is essential for editing tasks using diffusion models, yet existing methods produce non-Gaussian noise with poor editability due to the inaccuracy in early noising steps. We identify the root cause: a mathematical singularity that renders inversion fundamentally ill-posed. We propose Singularity Skipping Inversion of Diffusion Models (SSI-DM), which bypasses this singular region by adding small noise before standard inversion. This simple approach produces inverted noise with natural Gaussian properties while maintaining reconstruction fidelity. As a plug-and-play technique compatible with general diffusion models, our method achieves superior performance on public image datasets for reconstruction and interpolation tasks, providing a principled and efficient solution to diffusion model inversion.