Selective Frequency Restoration via Dual-Aware State Space Modeling
Abstract
Low-light image enhancement remains challenging due to the spatially heterogeneous nature of degradation, as different regions within the same image often require substantially different degrees of illumination correction and detail restoration. However, existing methods typically adopt spatially uniform processing, which is often inadequate for such diverse restoration demands. In this paper, we propose a frequency-aware and content-prioritized framework for low-light image enhancement. Specifically, the input image is decomposed into low- and high-frequency components via the discrete wavelet transform, enabling illumination-dominant and detail-dominant information to be restored separately. Building upon this decomposition, we introduce a Complexity-Brightness Dual-Aware State Space Group (CBDA-SSG), which performs selective patch-wise enhancement in a coarse-to-fine manner by prioritizing low-frequency patches with low brightness and high-frequency patches with high structural complexity. Furthermore, to mitigate the inconsistency between the two independently enhanced frequency branches, we propose an Inter-Frequency Attention Enhancement Module (IFAEM) that enables bidirectional cross-frequency refinement. By integrating content-prioritized restoration and cross-frequency interaction, the proposed framework achieves more targeted illumination adjustment and detail recovery. Extensive experiments on multiple benchmark datasets demonstrate that our method achieves superior visual quality and competitive quantitative performance compared with existing state-of-the-art approaches.