Training-Free Rate-Distortion-Perception Traversal With Diffusion

Yuhan Wang; Suzhi Bi; Ying-Jun Angela Zhang

cs.ITMar 4, 2026

Training-Free Rate-Distortion-Perception Traversal With Diffusion

Authors: Yuhan Wang, Suzhi Bi, Ying-Jun Angela Zhang

Organizations: Department of Information Engineering, The Chinese University of Hong Kong, Hong Kong · College of Electronic and Information Engineering, Shenzhen University, Shenzhen 518060, China

Abstract

The rate-distortion-perception (RDP) tradeoff characterizes the fundamental limits of lossy compression by jointly considering bitrate, reconstruction fidelity, and perceptual quality. While recent neural compression methods have improved perceptual performance, they typically operate at fixed points on the RDP surface, requiring retraining to target different tradeoffs. In this work, we propose a training-free framework that leverages pre-trained diffusion models to traverse the entire RDP surface. Our approach integrates a reverse channel coding (RCC) module with a novel score-scaled probability flow ODE decoder. We theoretically prove that the proposed diffusion decoder is optimal for the distortion-perception tradeoff under AWGN observations and that the overall framework with the RCC module achieves the optimal RDP function in the Gaussian case. Empirical results across multiple datasets demonstrate the framework's flexibility and effectiveness in navigating the ternary RDP tradeoff using pre-trained diffusion models. Our results establish a practical and theoretically grounded approach to adaptive, perception-aware compression.

Figures & tables

Fig. 1 : The proposed framework to traverse the RDP function using pre-trained diffusion models.

Fig. 2 : Information-theoretical RDP function for scalar Gaussian source (dashed line) and achieved rate, MSE, and W2 distance levels by our scheme (solid dots). (a) The RDP surface. (b) $R(D,P)$ function along DP planes. Different colors represent different rates.

Fig. 3 : Effects of controlling

t

and

\rho

on different metrics for the CIFAR-10 dataset. Distortion is quantified by MSE, and perception is measured by LPIPS and FID.

Fig. 4 : Rate-distortion-perception curves on the CIFAR-10 dataset. Distortion levels are quantified by MSE and perception levels are measured by LPIPS.

Fig. 5 : RDP tradeoff traversed by our proposed scheme on the Kodak and DIV2K datasets. We show the results obtained with Stable Diffusion (SD) 2.1 and the Flux model, respectively. More tradeoffs measured in different metrics (e.g., PSNR and FID) can be found in Appendix F-B2 .

	Param.	Enc. (s)	Dec. (s)	Total (s)
HiFiC [ 10 ]	181M	0.67	1.53	2.20
CDC [ 11 ]	53.8M	0.07	3.25	3.32
OneDC [ 15 ]	1.4B	0.31	0.43	0.74
StableCodec [ 17 ]	1.1B	0.27	0.47	0.74
Ours	950M	0.22-9.14	0.33-2.10	2.31-9.47
DDCM [ 22 ]	950M	37.76	37.91	75.67

TABLE I : Comparison of different methods in terms of model size and latency on Kodak dataset. Both our approach and DDCM utilize SD-2.1 as the backbone diffusion model. Notably, our framework is training-free . By leveraging a single pre-trained model, we can traverse a wide range of RDP tradeoffs, thereby significantly reducing training time and model storage requirements.

Fig. 6 : Samples from HiFiC, CDC, and our proposed schemes with different $\rho$ and $t$ .

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

		Rate Control	DP Control	Proved Optimality in Gaussian
Image Restoration	[ 30 ]	/	✓	✗
	[ 32 ]	/	✓	✓
	[ 31 ]	/	✓	✓
	Our ODE decoder	/	✓	✓
Lossy Compression	HiFiC [ 10 ]	✗	✗	✗
	CDC [ 11 ]	✗	✗	✗

Appendix

TABLE II: Comparison of related schemes in image restoration and lossy compression problems. Here the rate or DP control means the ability to adjust the rate or DP tradeoff with a single pre-trained model.

Fig. 7 : RDP curves on CIFAR-10 using different metrics.

Fig. 8 : Sample reconstructions on CIFAR-10 under varying $t$ and $\rho$ . Higher $\rho$ produces more vivid but less faithful images.

Fig. 9 : RDP metrics under

\rho=1

for SD1.5/2.1/XL and Flux.

Fig. 10 : Effect of controlling

t

and

\rho

on different metrics for the Kodak and DIV2K datasets. Stable Diffusion 2.1 and Flux depict different rate-distortion (R-D) and rate-perception (R-P) curves.

Fig. 11 : RDP curves for Kodak and DIV2K using PSNR vs. LPIPS under SD2.1 and Flux.

Fig. 12 : RDP curves for Kodak using MSE vs. FID under SD2.1 and Flux.

MSE ( $\times 10^{-3}$ ) /FID	$t$ values
MSE ( $\times 10^{-3}$ ) /FID	10	30	50	99	200	296	498
$\rho=0.83$	3.239/ 24.158	3.782/ 27.854	4.209/ 31.040	5.038/ 38.579	6.718/ 59.002	8.539/ 89.957	13.458/ 147.872
$\rho=0.85$	3.238/ 24.026	3.783/ 27.529	4.210/ 30.425	5.040/ 37.295	6.725/ 55.286	8.549/ 84.503	13.509/ 141.376
$\rho=0.88$	3.238/ 23.823	3.784/ 27.026	4.213/ 29.601	5.046/ 35.397	6.744/ 49.927	8.586/ 75.094	13.619/ 130.549
$\rho=0.90$	3.237/ 23.692	3.786/ 26.734	4.216/ 29.067	5.054/ 34.241	6.766/ 46.616	8.628/ 68.504	13.732/ 120.375
$\rho=0.92$	3.237/ 23.544	3.788/ 26.405	4.219/ 28.509	5.062/ 33.085	6.795/ 43.259	8.688/ 61.814	13.921/ 110.608

Appendix

TABLE III : MSE and FID values across different

\rho

and

t

values on Kodak dataset (SD 2.1)

Fig. 13 : Sample reconstructions on Kodak dataset under different $t$ and $\rho$ .

Fig. 14 : Sample reconstructions on DIV2K dataset under different $t$ and $\rho$ .

Fig. 15 : Sample reconstructions provided by Flux under different $t$ and $\rho$ .

Fig. 16 : Sample reconstructions with high resolution details under different $t$ and $\rho$ .

Fig. 17 : Sample reconstructions with high resolution details under different $t$ and $\rho$ .

Generative Image Compression

Generative image compression uses compact codes and learned generative priors to reconstruct plausible images at very low bitrates.

See all 35 papers

Dual-Constrained Diffusion Image Compression for Operational Rate-Distortion-Perception Optimization
Jun 11, 2026Sanxin Jiang, Jiro Katto, Heming SunLearned Image Compression Rate-Distortion Optimization
Department of Information Engineering Shanghai University of Electric Power Shanghai, China · Department of Computer Science and Communication Engineering Waseda University Tokyo, Japan · Faculty of Engineering Institute of Science Tokyo Tokyo, Japan
Rate-Distortion-Perception Theory: Redefining the Fundamental Limits of Information Representation
Jul 19, 2026Photios A. Stavrou, Giuseppe Serra, Marios KountourisLearned Image Compression Rate-Distortion Theory
Advances in Diffusion-Based Generative Compression
Jan 26, 2026Yibo Yang, Stephan MandtLearned Image Compression Generative Image Compression
Chan Zuckerberg Biohub · University of California Irvine

Jun 11, 2026J/K move · Enter open · S save

Dual-Constrained Diffusion Image Compression for Operational Rate-Distortion-Perception Optimization

Sanxin Jiang, Jiro Katto, Heming Sun

Department of Information Engineering Shanghai University of Electric Power Shanghai, China · Department of Computer Science and Communication Engineering Waseda University Tokyo, Japan · Faculty of Engineering Institute of Science Tokyo Tokyo, Japan

Learned Image Compression Rate-Distortion Optimization Distortion-Perception Tradeoff Generative Image Compression Diffusion Models

Open paper AskPDF arXiv

The rate-distortion-perception (RDP) trade-off extends classical rate--distortion theory by imposing a distributional constraint on reconstructions, providing a unified framework for neural image compression that jointly governs fidelity and perceptual realism. While prior work achieves near-optimal rate--perception trade-offs, practical frameworks explicitly realizing the full RDP surface remain scarce, primarily due to the difficulty of introducing common randomness at the decoder. We propose DCIC (Dual-Constrained Diffusion Image Compression), which integrates a learned codec with a diffusion-based decoder governed by joint distortion and idempotence constraints. The distortion constraint bounds reconstruction fidelity relative to the base codec output; the idempotence constraint -- requiring that re-encoding the restored image recovers the base codec reconstruction -- serves as a tractable surrogate for the distributional perception requirement. Together, they steer the reverse denoising process via iterative optimization with consistent noise injection, realizing common randomness without additional rate overhead. At fixed rate, dual attenuation factors

(K_D, K_P)

jointly navigate the Pareto frontier of the distortion-perception plane, enabling continuously adjustable fidelity-realism trade-offs from a single bitstream. DCIC

_{RD}

(

K_P{=}0

) and DCIC

_{RP}

(

K_D{=}0

) arise as boundary curves, with DCIC

_{RDP}

(

K_D = K_P=1

) realizing the optimal interior operating point. Experiments on CelebA-HQ, CLIC2020, and ImageNet-1K across CNN, Transformer, and hybrid architectures confirm that DCIC

_{RDP}

achieves superior BD-PSNR over all perceptual codecs, while DCIC

_{RP}

matches dedicated perception-oriented methods in BD-FID, validating the practical value of full RDP surface navigation.

Abstract

Generative Image Compression

Explore similar work

Dual-Constrained Diffusion Image Compression for Operational Rate-Distortion-Perception Optimization

Rate-Distortion-Perception Theory: Redefining the Fundamental Limits of Information Representation

Advances in Diffusion-Based Generative Compression