arxiv:2606.16396

SP^3: Spherical Priors for Plug-and-Play Restoration

Published on Jun 15

· Submitted by

Sean Man on Jun 16

Upvote

Authors:

Sean Man ,

Matan Kleiner ,

Abstract

SP³ uses spherical encoders as generative priors to accelerate maximum a posteriori image restoration, enabling fast convergence and high-quality results through structured latent space projections.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

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 3-630times faster.

View arXiv page View PDF Project page GitHub 2 Add to collection

Community

seanman

Paper author Paper submitter about 12 hours ago

Hi everyone, excited to share our new paper on Plug-and-Play restoration using Spherical Priors.

The main idea is to replace the usual denoiser/diffusion prior in PnP restoration with a Spherical Encoder prior. SP³ returns a natural-looking image after each iteration, and subsequent iterations improve the quality and consistency. This "anytime" restoration behavior is a key advantage over diffusion-based methods that require choosing the number of denoising steps in advance, and running all the way to the end.

In our experiments, SP³ reaches perceptual quality comparable to zero-shot diffusion and flow priors while being 3–630× faster across restoration tasks.

Happy to answer questions about the algorithm, results, or the sphere-prior design.