SP$^3$: Spherical Priors for Plug-and-Play Restoration
2026-06-15 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present SP³, a new method to speed up image restoration by using Spherical Encoders instead of traditional denoisers. Their approach projects corrupted images onto a space that represents natural images, helping to clean them up more reliably. By alternating this projection with a straightforward step ensuring the result matches observed data, their method converges quickly without complex calculations. Tests show SP³ produces sharp, high-quality images much faster than other leading zero-shot restoration techniques.
image restorationmaximum a posteriori (MAP)Spherical Encodersgenerative priorsproximal operatorHalf-Quadratic Splittingzero-shot learninglatent spacedenoisingimage manifold
Authors
Sean Man, Ron Raphaeli, Matan Kleiner, Or Ronai
Abstract
In this paper, we introduce SP$^3$, a novel Plug-and-Play algorithm that accelerates maximum a posteriori image restoration by replacing denoisers with Spherical Encoders (SE) as generative priors. SP$^3$ approximates the intractable proximal prior step by utilizing the SE tightly structured latent space as a robust projection onto the natural image manifold. Alternating this projection with a closed-form data-consistency step, via Half-Quadratic Splitting, achieves stable convergence without requiring gradient computation during inference. This unique formulation unlocks "anytime" restoration capabilities, producing sharp, plausible images from the first iteration. Evaluations across a variety of image restoration tasks demonstrate that SP$^3$ achieves perceptual quality comparable to state-of-the-art zero-shot diffusion and flow methods while being $3$-$630\times$ faster.