Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution

TL;DR

This paper introduces a diffusion-based generative image compression method that models the compression process as a stochastic differential equation, enabling photo-realistic reconstructions with smooth rate control and minimal sampling steps.

Contribution

It presents a novel diffusion framework for generative image compression that directly models the compression process as an SDE, improving reconstruction quality and rate flexibility.

Findings

Outperforms existing GIC methods on benchmark datasets

Achieves high perceptual and statistical quality with few sampling steps

Enables smooth rate adjustment in image compression

Abstract

While learned image compression (LIC) focuses on efficient data transmission, generative image compression (GIC) extends this framework by integrating generative modeling to produce photo-realistic reconstructed images. In this paper, we propose a novel diffusion-based generative modeling framework tailored for generative image compression. Unlike prior diffusion-based approaches that indirectly exploit diffusion modeling, we reinterpret the compression process itself as a forward diffusion path governed by stochastic differential equations (SDEs). A reverse neural network is trained to reconstruct images by reversing the compression process directly, without requiring Gaussian noise initialization. This approach achieves smooth rate adjustment and photo-realistic reconstructions with only a minimal number of sampling steps. Extensive experiments on benchmark datasets demonstrate that our method outperforms existing generative image compression approaches across a range of metrics, including perceptual distortion, statistical fidelity, and no-reference quality assessments.