Diffusion Plug-in Priors for Self-Supervised Ultrasound Video Restoration

Chu Chen; Mingduo He; Yining Li; Yinhao Wu; Wei Tang; Yiwen Luo; Yixuan Chen; Pasquale Cascarano

cs.CVNov 3, 2026

Diffusion Plug-in Priors for Self-Supervised Ultrasound Video Restoration

Authors: Chu Chen, Mingduo He, Yining Li, Yinhao Wu, Wei Tang, Yiwen Luo, Yixuan Chen, Pasquale Cascarano,

Organizations: Department of Engineering, University of Cambridge, Cambridge, UK · Department of Biomedical Engineering, School of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China · Department of Intelligent Medical Engineering, School of Medicine, Nankai University, Tianjin, China · University of Texas at Arlington, Arlington, TX, USA · Department of Engineering Science, University of Oxford, Oxford, UK · Oxford Suzhou Centre for Advanced Research, Suzhou, China · Department of Electrical Engineering, City University of Hong Kong, Hong Kong SAR · Department of the Arts, University of Bologna, Bologna, Italy · Department of Data Science, City University of Hong Kong (Dongguan), Dongguan, China · City University of Hong Kong, Hong Kong SAR

Abstract

Ultrasound videos are limited by hardware-constrained spatial resolution and signal-dependent speckle, while paired low- and high-resolution acquisitions for supervised restoration are rarely available. We aim to jointly super-resolve and despeckle ultrasound videos without paired training data or network retraining. We propose UltraDiff, which uses a frozen pre-trained diffusion model (DM) as a plug-in prior and restores each frame by optimizing the latent seed of the DM sampler under a range-constrained maximum a posteriori formulation. We derive a speckle-aware data-fidelity term from a Rayleigh scale-family model and characterize its relative-residual and variance-adaptive weighting properties. We further introduce a stabilized fidelity that is bounded below and strictly convex in log intensity for positive observations in its unclamped form. A total-variation regularizer complements the DM prior, and a temporal seed update warm-starts each frame from the optimized seed of its predecessor. On EchoNet-LVH videos with simulated speckle and additive noise, UltraDiff with a natural-image prior improves PSNR by 1.24,dB over the strongest baseline, and ultrasound-related priors improve it further. On real clinical echocardiography videos, UltraDiff reduces the mean absolute error of zero-shot ejection fraction (EF) prediction from 7.31 to 6.37 EF points relative to DUP, the strongest baseline. Combining a speckle-aware fidelity with frozen diffusion priors enables ground-truth-free ultrasound video restoration that benefits downstream measurement. UltraDiff needs neither paired data nor retraining and is not tied to a specific diffusion checkpoint. Code associated with this work will be made publicly available upon acceptance.

Figures & tables

Figure 1: Motivating comparison of ultrasound video restoration. We compare reconstructed frames from LR ultrasound sequences with HR references and representative methods, including EDSR, ZSSR-Cycle, and UltraDiff. The zoom-in regions highlight structural details and tissue texture: UltraDiff suppresses noise while better preserving anatomical structures and the tissue echotexture.

Figure 2: Overview of the UltraDiff framework. A frozen pre-trained diffusion model maps latent seeds to HR estimates, which are degraded back to the LR space and optimized against the observed LR frames. The optimized seeds are propagated and perturbed across consecutive frames to enable video-adaptive restoration without diffusion model retraining.

Figure 3: Qualitative comparison for 2 $\times$ ultrasound video super-resolution. The left panel shows the HR reference and the simulated LR input. The top row compares SR outputs from different methods, with PSNR /SSIM reported below each result, and the bottom row shows the corresponding error maps with zoomed regions for structural inspection.

Figure 4: Robustness analysis under increasing mixed-noise levels from 0.02 to 0.10. Box-and-violin plots report PSNR and SSIM across different methods. UltraDiff maintains higher medians and more concentrated distributions, indicating stronger restoration stability under heavier degradation.

Method	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	NIQE $\downarrow$	PIQE $\downarrow$	TOPIQ-NR $\uparrow$
Bicubic	21.76	0.5323	0.1721	11.0904	43.7747	0.4173
DiffVSR [ 32 ]	23.80	0.7395	0.1011	9.4150	47.1095	0.4651
DUP [ 29 ]	23.67	0.8096	0.1665	10.0552	48.7218	0.3836
EDSR [ 18 ]	21.87	0.7573	0.2193	17.3222	57.1770	0.4848
ZSSR-Cycle [ 25 ]	22.91	0.7894	0.2027	17.5018	50.1141	0.4431
UltraDiff-DFUI	25.27	0.8229	0.1843	7.8461	40.4669	0.4624

Table 1: Quantitative restoration performance on EchoNet-LVH. UltraDiff variants use different pre-trained diffusion priors. The best, second-best, and third-best results are highlighted in red , orange , and blue , respectively.

Variations	PSNR↑	SSIM↑	LPIPS↓	NIQE↓	PIQE↓
MSE fidelity	24.42	0.8087	0.1233	9.7698	38.0033
w/o TV	24.47	0.8076	0.1210	9.4159	38.6782
w/o SUP	24.13	0.7636	0.2934	11.2284	76.2607
UltraDiff	25.04	0.8209	0.1101	9.1370	36.2833

Table 2: Ablation study of UltraDiff components on restoration quality. Removing each component degrades performance.

Method	MAE $\downarrow$	Acc@ $\pm$ 5 $\uparrow$	Recall@5 ±5 $\uparrow$
Bicubic	11.00	28.00	67.00
EDSR	10.60	26.00	58.00
ZSSR-Cycle	8.07	37.00	72.00
DiffVSR	8.35	36.00	60.00
DUP	7.31	44.00	72.00
UltraDiff	6.37	53.00	74.00

Table 3: Downstream zero-shot EF prediction on the 100-video EchoNet-Dynamic subset after SR. MAE (EF %), Acc@

\pm

5, and Recall@5 ±5 are reported. UltraDiff denotes the default CelebA-HQ setting.

Figure 5: Accuracy-error threshold curves for downstream EF prediction on the 100-video EchoNet-Dynamic subset.

\mathrm{Acc@}k

measures the percentage of cases whose absolute EF error is within

k

EF points. Higher curves indicate more reliable EF estimation across different error tolerances.

Panel of Figure 6 — Figure 6: Bland–Altman analysis of downstream EF prediction on the 100-video EchoNet-Dynamic subset. The solid line denotes the mean prediction bias, and the dashed lines denote the 95% limits of agreement ( $\mathrm{bias}\pm 1.96\,\mathrm{SD}$ ). UltraDiff shows the smallest bias and the narrowest agreement range among the compared SR methods.

Video Restoration

Video restoration reconstructs clearer, more faithful, temporally consistent footage from blur, noise, compression, weather, and other degradations.

See all 61 papers

IRSDE-Despeckle: A Physics-Grounded Diffusion Model for Generalizable Ultrasound Despeckling
Feb 26, 2026Shuoqi Chen, Yujia Wu, Geoffrey P. LukeUltrasound Imaging Image Denoising
Thayer School of Engineering, Dartmouth College, Hanover, NH, USA
Null-Space Diffusion Restoration with Adaptive Uncertainty-Guided Fusion for Ultrasound Speckle Reduction
Aug 30, 2026Juneyong Lee, Jaeyoung ChoiDiffusion Models for Image Restoration Ultrasound Imaging
Department of Computer Science and Engineering, Hankuk University of Foreign Studies, Yongin-si 17035, Republic of Korea
Mask2Restore: Self-Supervised Ultrasound Despeckling via Inpainting
Sep 26, 2026Xuesong Li, Yingtai Xu, Zhongliang Jiang +2Image Restoration Image Denoising
Computer Aided Medical Procedures (CAMP), Technical University of Munich, Munich, Germany · Munich Center for Machine Learning (MCML), Munich, Germany · Department of Mechanical Engineering, The University of Hong Kong, Hong Kong SAR, China

Feb 26, 2026J/K move · Enter open · S save

IRSDE-Despeckle: A Physics-Grounded Diffusion Model for Generalizable Ultrasound Despeckling

Shuoqi Chen, Yujia Wu, Geoffrey P. Luke

Thayer School of Engineering, Dartmouth College, Hanover, NH, USA

Ultrasound Imaging Image Denoising Diffusion Models for Image Restoration Physics-Guided Image Restoration

Open paper AskPDF arXiv

Ultrasound imaging is widely used for real-time, noninvasive diagnosis, but speckle and related artifacts reduce image quality and can hinder interpretation. We present a diffusion-based ultrasound despeckling method built on the Image Restoration Stochastic Differential Equations framework. To enable supervised training, we curate large paired datasets by simulating ultrasound images from speckle-free magnetic resonance images using the Matlab UltraSound Toolbox. The proposed model reconstructs speckle-suppressed images while preserving anatomically meaningful edges and contrast. On a held-out simulated test set, our approach consistently outperforms classical filters and recent learning-based despeckling baselines. We quantify prediction uncertainty via cross-model variance and show that higher uncertainty correlates with higher reconstruction error, providing a practical indicator of difficult or failure-prone regions. Finally, we evaluate sensitivity to simulation probe settings and observe domain shift, motivating diversified training and adaptation for robust clinical deployment.

Abstract

Video Restoration

Explore similar work

IRSDE-Despeckle: A Physics-Grounded Diffusion Model for Generalizable Ultrasound Despeckling

Null-Space Diffusion Restoration with Adaptive Uncertainty-Guided Fusion for Ultrasound Speckle Reduction

Mask2Restore: Self-Supervised Ultrasound Despeckling via Inpainting