PRE-Mamba: A 4D State Space Model for Ultra-High-Frequent Event Camera Deraining

TL;DR

PRE-Mamba is a novel 4D state space model that significantly improves event camera deraining by capturing complex rain dynamics with high efficiency and generalization across diverse weather conditions.

Contribution

It introduces a 4D event cloud representation, a Spatio-Temporal Decoupling and Fusion module, and a Multi-Scale State Space Model for superior deraining performance.

Findings

Achieves state-of-the-art deraining accuracy with 0.95 SR and 0.91 NR.

Operates with only 0.26M parameters and 0.4s/M events.

Generalizes well across rain intensities, viewpoints, and snowy conditions.

Abstract

Event cameras excel in high temporal resolution and dynamic range but suffer from dense noise in rainy conditions. Existing event deraining methods face trade-offs between temporal precision, deraining effectiveness, and computational efficiency. In this paper, we propose PRE-Mamba, a novel point-based event camera deraining framework that fully exploits the spatiotemporal characteristics of raw event and rain. Our framework introduces a 4D event cloud representation that integrates dual temporal scales to preserve high temporal precision, a Spatio-Temporal Decoupling and Fusion module (STDF) that enhances deraining capability by enabling shallow decoupling and interaction of temporal and spatial information, and a Multi-Scale State Space Model (MS3M) that captures deeper rain dynamics across dual-temporal and multi-spatial scales with linear computational complexity. Enhanced by frequency-domain regularization, PRE-Mamba achieves superior performance (0.95 SR, 0.91 NR, and 0.4s/M events) with only 0.26M parameters on EventRain-27K, a comprehensive dataset with labeled synthetic and real-world sequences. Moreover, our method generalizes well across varying rain intensities, viewpoints, and even snowy conditions.