PnP-CM: Consistency Models as Plug-and-Play Priors for Inverse Problems

Figure 1. Representative results of our method (PnP-CM) across a diverse set of inverse problems. Left to right: linear inverse problems (Gaussian deblurring, super-resolution, inpainting) and nonlinear inverse problems (JPEG artifact removal, nonlinear deblurring, phase retrieval), all with additive Gaussian noise σ = 0.05, followed by MRI reconstruction with inherently noisy measurements. Top to bottom: reference, degraded measurement, and PnP-CM reconstruction.