Effective Real Image Editing with Accelerated Iterative Diffusion Inversion

TL;DR

This paper introduces AIDI, an accelerated iterative diffusion inversion method that enhances real image editing by improving reconstruction accuracy and robustness with minimal computational overhead, enabling efficient editing with fewer diffusion steps.

Contribution

The paper presents a novel blended guidance technique and an accelerated inversion process that significantly improves the stability and efficiency of real image editing using diffusion models.

Findings

AIDI achieves higher reconstruction accuracy than existing methods.

The approach is more robust in fast editing regimes with 10-20 diffusion steps.

It requires minimal additional computational resources.

Abstract

Despite all recent progress, it is still challenging to edit and manipulate natural images with modern generative models. When using Generative Adversarial Network (GAN), one major hurdle is in the inversion process mapping a real image to its corresponding noise vector in the latent space, since its necessary to be able to reconstruct an image to edit its contents. Likewise for Denoising Diffusion Implicit Models (DDIM), the linearization assumption in each inversion step makes the whole deterministic inversion process unreliable. Existing approaches that have tackled the problem of inversion stability often incur in significant trade-offs in computational efficiency. In this work we propose an Accelerated Iterative Diffusion Inversion method, dubbed AIDI, that significantly improves reconstruction accuracy with minimal additional overhead in space and time complexity. By using a novel blended guidance technique, we show that effective results can be obtained on a large range of image editing tasks without large classifier-free guidance in inversion. Furthermore, when compared with other diffusion inversion based works, our proposed process is shown to be more robust for fast image editing in the 10 and 20 diffusion steps' regimes.