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

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Han, Minghao, You, Weiyi, Zhang, Jinhua, Zhang, Leheng, Zhu, Ce, Gu, Shuhang
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912397645578240
author Han, Minghao
You, Weiyi
Zhang, Jinhua
Zhang, Leheng
Zhu, Ce
Gu, Shuhang
author_facet Han, Minghao
You, Weiyi
Zhang, Jinhua
Zhang, Leheng
Zhu, Ce
Gu, Shuhang
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20984
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution
Han, Minghao
You, Weiyi
Zhang, Jinhua
Zhang, Leheng
Zhu, Ce
Gu, Shuhang
Image and Video Processing
Computer Vision and Pattern Recognition
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.
title Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.20984