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

2025-05-27

AI Summary

Existing generative image compression (GIC) methods suffer from reliance on Gaussian noise initialization, discrete bit-rate control, and a fundamental trade-off between reconstruction fidelity and coding efficiency. To address these limitations, this paper proposes an end-to-end differentiable compression framework grounded in stochastic differential equations (SDEs). Its core innovations are threefold: (1) modeling image compression as a controllable forward diffusion process—eliminating noise-based initialization and enabling direct image reconstruction via a lightweight learnable reverse network; (2) introducing rate-variable feature-distribution gradient estimation for continuous, fine-grained bit-rate adaptation; and (3) adopting no-reference, quality-driven training to jointly optimize rate–distortion–perception performance. Extensive experiments on standard benchmarks demonstrate that our method consistently outperforms state-of-the-art generative compressors across perceptual distortion (LPIPS), statistical fidelity (FID), and no-reference quality metrics (CLIPIQ, MANIQA).

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Data Mining & Knowledge Management: Data Compression

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

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.

Problem

• Proposes diffusion-based generative modeling for image compression
• Reinterprets compression as forward diffusion using SDEs
• Achieves photo-realistic reconstructions with minimal sampling steps

Innovation

• Diffusion-based generative modeling for compression
• Reinterprets compression as forward diffusion path
• Reverse neural network reconstructs without noise initialization