Learning to Remove Lens Flare in Event Camera

TL;DR

This paper introduces E-Deflare, a comprehensive framework for removing lens flare from event camera data, including a physics-based model, a large-scale benchmark, and a neural network that achieves state-of-the-art results.

Contribution

It provides the first systematic approach to lens flare removal in event cameras, including a theoretical model, a large benchmark dataset, and a specialized neural network.

Findings

E-DeflareNet achieves state-of-the-art restoration performance.

The benchmark datasets enable robust training and evaluation.

The approach benefits downstream vision tasks.

Abstract

Event cameras have the potential to revolutionize vision systems with their high temporal resolution and dynamic range, yet they remain susceptible to lens flare, a fundamental optical artifact that causes severe degradation. In event streams, this optical artifact forms a complex, spatio-temporal distortion that has been largely overlooked. We present E-Deflare, the first systematic framework for removing lens flare from event camera data. We first establish the theoretical foundation by deriving a physics-grounded forward model of the non-linear suppression mechanism. This insight enables the creation of the E-Deflare Benchmark, a comprehensive resource featuring a large-scale simulated training set, E-Flare-2.7K, and the first-ever paired real-world test set, E-Flare-R, captured by our novel optical system. Empowered by this benchmark, we design E-DeflareNet, which achieves state-of-the-art restoration performance. Extensive experiments validate our approach and demonstrate clear benefits for downstream tasks. Code and datasets are publicly available.