Recent advances in low-light image enhancement have leveraged various physical priors to significantly improve visual quality. Among them, the Signal-to-Noise Ratio (SNR) priors have demonstrated strong potential by modeling long-range dependencies in low-SNR regions and short-range correlations in high-SNR areas. However, the conventional use of SNR priors solely in the spatial domain overlooks their potential benefits in the frequency domain, which in turn may limit their overall effectiveness. In this paper, we revisit the existing SNR priors and first identify their dual property, which simultaneously reflects both the lightness degrees and structural information. Inspired by the dual property and necessity analysis, we propose the SNR-based Attention Solver to jointly modulate amplitude (lightness) and phase (structure) components in the frequency domain. To further enhance selectivity, we introduce a Selective Frequency Domain-based FFN, which adaptively filters and retains valuable frequency features. Additionally, we propose the Frequency-based Dual-lightness Fusion Module to extend the applicability of lightness enhancement across a wide range of lighting conditions. Extensive experiments on seven benchmarks show our superiority among the state-of-the-art methods.