D2-Mamba: Dual-Scale Fusion and Dual-Path Scanning with SSMs for Shadow Removal

TL;DR

This paper introduces D2-Mamba, a novel shadow removal network that uses dual-scale fusion and dual-path scanning to effectively leverage contextual cues and adapt to region-specific transformations, outperforming existing methods.

Contribution

The paper proposes a new Mamba-based network with dual-scale fusion and dual-path scanning for improved shadow removal performance.

Findings

Outperforms state-of-the-art shadow removal methods on benchmark datasets.

Effectively reduces boundary artifacts and enhances structural continuity.

Improves fine-grained region modeling through adaptive scanning strategies.

Abstract

Shadow removal aims to restore images that are partially degraded by shadows, where the degradation is spatially localized and non-uniform. Unlike general restoration tasks that assume global degradation, shadow removal can leverage abundant information from non-shadow regions for guidance. However, the transformation required to correct shadowed areas often differs significantly from that of well-lit regions, making it challenging to apply uniform correction strategies. This necessitates the effective integration of non-local contextual cues and adaptive modeling of region-specific transformations. To this end, we propose a novel Mamba-based network featuring dual-scale fusion and dual-path scanning to selectively propagate contextual information based on transformation similarity across regions. Specifically, the proposed Dual-Scale Fusion Mamba Block (DFMB) enhances multi-scale feature representation by fusing original features with low-resolution features, effectively reducing boundary artifacts. The Dual-Path Mamba Group (DPMG) captures global features via horizontal scanning and incorporates a mask-aware adaptive scanning strategy, which improves structural continuity and fine-grained region modeling. Experimental results demonstrate that our method significantly outperforms existing state-of-the-art approaches on shadow removal benchmarks.