Lune

ICLR2025Top-tier venue

Once-for-All: Controllable Generative Image Compression with Dynamic Granularity Adaptation

Anqi Li, Feng Li, Yuxi Liu, Runmin Cong, Yao Zhao, Huihui Bai

Abstract

Although recent generative image compression methods have demonstrated impressive potential in optimizing the rate-distortion-perception trade-off, they still face the critical challenge of flexible rate adaptation to diverse compression necessities and scenarios. To overcome this challenge, this paper proposes a Controllable Generative Image Compression framework, termed Control-GIC, the first capable of fine-grained bitrate adaptation across a broad spectrum while ensuring high-fidelity and generality compression. Control-GIC is grounded in a VQGAN framework that encodes an image as a sequence of variable-length codes (i.e. VQ-indices), which can be losslessly compressed and exhibits a direct positive correlation with bitrates. Drawing inspiration from the classical coding principle, we correlate the information density of local image patches with their granular representations. Hence, we can flexibly determine a proper allocation of granularity for the patches to achieve dynamic adjustment for VQ-indices, resulting in desirable compression rates. We further develop a probabilistic conditional decoder capable of retrieving historic encoded multi-granularity representations according to transmitted codes, and then reconstruct hierarchical granular features in the formalization of conditional probability, enabling more informative aggregation to improve reconstruction realism. Our experiments show that Control-GIC allows highly flexible and controllable bitrate adaptation where the results demonstrate its superior performance over recent state-of-the-art methods. Code is available at https://github.com/lianqi1008/Control-GIC.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

2 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • ControlVideo: Training-free Controllable Text-to-video Generation cites Denoising Diffusion Probabilistic Models
  • Efficient Learned Image Compression without Entropy Coding cites Generative Adversarial Networks for Extreme Learned Image Compression
  • Efficient Learned Image Compression without Entropy Coding cites Denoising Diffusion Probabilistic Models
  • Efficient Learned Image Compression without Entropy Coding cites High-Fidelity Generative Image Compression
  • Differentiable Vector Quantization for Rate-Distortion Optimization of Generative Image Compression cites Lossy Image Compression with Conditional Diffusion Models
  • Differentiable Vector Quantization for Rate-Distortion Optimization of Generative Image Compression cites Generative Adversarial Networks for Extreme Learned Image Compression
  • Differentiable Vector Quantization for Rate-Distortion Optimization of Generative Image Compression cites Denoising Diffusion Probabilistic Models
  • Differentiable Vector Quantization for Rate-Distortion Optimization of Generative Image Compression cites High-Fidelity Generative Image Compression
  • Lossy Image Compression with Conditional Diffusion Models cites Generative Adversarial Networks for Extreme Learned Image Compression
  • Lossy Image Compression with Conditional Diffusion Models cites Denoising Diffusion Probabilistic Models
  • Lossy Image Compression with Conditional Diffusion Models cites High-Fidelity Generative Image Compression
  • High-Fidelity Generative Image Compression cites Generative Adversarial Networks for Extreme Learned Image Compression
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers2

  1. Differentiable Vector Quantization for Rate-Distortion Optimization of Generative Image CompressionShiyin Jiang, Wei Long, Minghao Han, Zhenghao Chen et al.CVPR 2026 · 3 citations
  2. Efficient Learned Image Compression without Entropy CodingHao Cao, Wenqi Guo, Zhijin Qin, Jungong HanICML 2026

Builds on25

  1. Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 citations
  2. High-Fidelity Generative Image CompressionFabian Mentzer, George Toderici, Michael Tschannen, Eirikur AgustssonNeurIPS 2020 · 675 citations
  3. Generative Adversarial Networks for Extreme Learned Image CompressionEirikur Agustsson, Michael Tschannen, Fabian Mentzer, Radu Timofte et al.ICCV 2019 · 648 citations
  4. ControlVideo: Training-free Controllable Text-to-video GenerationYabo Zhang, Yuxiang Wei, Dongsheng Jiang, Xiaopeng Zhang et al.ICLR 2024 · 359 citations
  5. Lossy Image Compression with Conditional Diffusion ModelsRuihan Yang, Stephan MandtNeurIPS 2023 · 268 citations

Related papers

  1. Variable Rate Deep Image Compression With a Conditional AutoencoderYoojin Choi, Mostafa El-Khamy, Jungwon LeeICCV 2019 · 265 citations
  2. Decouple Distortion from Perception: Region Adaptive Diffusion for Extreme-low Bitrate Perception Image CompressionJinchang Xu, Shaokang Wang, Jintao Chen, Zhe Li et al.CVPR 2025
  3. MoVQ: Modulating Quantized Vectors for High-Fidelity Image GenerationChuanxia Zheng, Tung-Long Vuong, Jianfei Cai, Dinh PhungNeurIPS 2022 · 156 citations
  4. Generative Latent Coding for Ultra-Low Bitrate Image CompressionZhaoyang Jia, Jiahao Li, Bin Li, Houqiang Li et al.CVPR 2024
  5. Improving Statistical Fidelity for Neural Image Compression with Implicit Local Likelihood ModelsMatthew J. Muckley, Alaaeldin El-Nouby, Karen Ullrich, Hervé Jégou et al.ICML 2023 · 114 citations
Dusk over the sea between two cliffs drawn in fine vertical lines