I developed a selection procedure for image formation and solver hyperparameters for image restoration tasks, which removes the need for hand tuning. The method reframes image restoration so that a solver predicts deliberately held-out pixels, allowing parameter choices to be evaluated using single-image, pixel-wise hold-out cross-validation.

Paper Code

Abstract

Historically, image formation parameters and hyperparameters have been tuned by hand for image restoration due to the lack of a broadly applicable selection framework. This has substantially limited the application of state-of-the-art image processing methods. We show that recasting the problem of image restoration to include the additional task of inpainting pixels permits an automatic hyperparameter selection procedure for image restoration through a pixel-wise hold-out cross-validation. This procedure is algorithm-agnostic and works on single images without any prior information. We show that the estimator in the procedure is affinely unbiased under very general settings and can be used for model selection. We provide a reference implementation to demonstrate how to extend existing methods to support inpainting and present empirical results. We hope that this eliminates manual tuning in image restoration pipelines.

Reference

@inproceedings{tierney2025single,
  author = {Tierney, Stephen},
  title = {Single Image Blind Parameter Recovery},
  booktitle = {2025 Digital Image Computing: Techniques and Applications (DICTA)},
  year = {2025},
  pages = {1--8},
  doi = {10.1109/DICTA68720.2025.11302445},
  url = {https://ieeexplore.ieee.org/abstract/document/11302445}
}
Image from paper 'Single Image Blind Parameter Recovery' DICTA poster