Abstract
Diffusion bridges (DB) have emerged as a promising alternative to diffusion models for imaging inverse problems, achieving faster sampling by directly bridging low- and high-quality image distributions. While incorporating measurement consistency has been shown to improve performance, existing DB methods fail to maintain this consistency in blind inverse problems, where the forward model is unknown. To address this limitation, we introduce ADOBI (Adaptive Diffusion Bridge for Inverse Problems), a novel framework that adaptively calibrates the unknown forward model to enforce measurement consistency throughout sampling iterations. Our adaptation strategy allows ADOBI to achieve high-quality parallel magnetic resonance imaging (PMRI) reconstruction in only 5–10 steps. Our numerical results show that ADOBI consistently delivers state-of-the-art performance, and further advances the Pareto frontier for the perception-distortion trade-off.
Adaptive Diffusion Bridge (ADOBI)
Figure 1: ADOBI is a novel diffusion bridge method for imaging inverse problems that adapts unknown forward models to enforce measurement consistency.
ADOBI for PMRI
Figure 2: Visual comparison of ADOBI with baseline methods on PMRI. The top row shows results for 4x accelerated PMRI data collection while the bottom row shows those 8x. Error maps and zoomed-in areas highlight differences. Note how ADOBI provides the best visual and quantitative performance in both settings
Perception-Distortion Tradeoff of ADOBI
Figure 3: Perception and distortion performance comparison at 8x acceleration across different NFEs. Left: NFE vs. PSNR; Right: NFE vs. LPIPS. ADOBI consistently outperforms baselines. Our method outperforms baseline methods across most NFE settings, achieving superior perception and distortion performance.
Uncertainty Quantification of ADOBI
Figure 4: Uncertainty quantification results on 4x (top row) and 8x (bottom row) acceleration PMRI images. Absolute error to ground truth corresponds to conditional mean, and the variance is calculated by pixel-wise standard deviation. Ill-posed nature of the task has a direct effect on the diversity of generated images, and variance is heavily correlated to reconstruction errors.