Scale Equivariance Regularization and Feature Lifting in High Dynamic Range Modulo Imaging

TL;DR

This paper introduces a learning-based HDR reconstruction method using scale-equivariant regularization and feature lifting, significantly improving the accuracy of recovering high dynamic range images from modulo measurements.

Contribution

It proposes a novel HDR restoration framework that combines scale-equivariant regularization with a feature lifting input design for better artifact discrimination.

Findings

Achieves state-of-the-art HDR reconstruction performance.

Enhances the network's ability to distinguish true image structures from wrap artifacts.

Improves perceptual and linear HDR quality metrics.

Abstract

Modulo imaging enables high dynamic range (HDR) acquisition by cyclically wrapping saturated intensities, but accurate reconstruction remains challenging due to ambiguities between natural image edges and artificial wrap discontinuities. This work proposes a learning-based HDR restoration framework that incorporates two key strategies: (i) a scale-equivariant regularization that enforces consistency under exposure variations, and (ii) a feature lifting input design combining the raw modulo image, wrapped finite differences, and a closed-form initialization. Together, these components enhance the network's ability to distinguish true structure from wrapping artifacts, yielding state-of-the-art performance across perceptual and linear HDR quality metrics.