Accelerated Image-Aware Generative Diffusion Modeling

TL;DR

This paper introduces a novel diffusion model that accelerates image generation by reducing diffusion steps, employing an autoencoder for learning diffusion coefficients, and enabling single-run reverse diffusion, thus significantly speeding up image synthesis.

Contribution

The paper presents a new diffusion model with exponentially decaying SNR, a parallel data-driven reverse process, and eliminates MCMC correction, achieving faster image generation without quality loss.

Findings

Reduces diffusion steps from ~1000 to 200-500

Enables single-run reverse diffusion process

Maintains high image quality and diversity

Abstract

We propose in this paper an analytically new construct of a diffusion model whose drift and diffusion parameters yield an exponentially time-decaying Signal to Noise Ratio in the forward process. In reverse, the construct cleverly carries out the learning of the diffusion coefficients on the structure of clean images using an autoencoder. The proposed methodology significantly accelerates the diffusion process, reducing the required diffusion time steps from around 1000 seen in conventional models to 200-500 without compromising image quality in the reverse-time diffusion. In a departure from conventional models which typically use time-consuming multiple runs, we introduce a parallel data-driven model to generate a reverse-time diffusion trajectory in a single run of the model. The resulting collective block-sequential generative model eliminates the need for MCMC-based sub-sampling correction for safeguarding and improving image quality, to further improve the acceleration of image generation. Collectively, these advancements yield a generative model that is an order of magnitude faster than conventional approaches, while maintaining high fidelity and diversity in generated images, hence promising widespread applicability in rapid image synthesis tasks.