Event-Driven Detail Enhancement of Low-Light Images via Frequency-Domain Fusion

This study introduces an event-driven detail enhancement framework that selectively incorporates event-based features into the high-frequency components of low-light images. The approach uses the discrete Fourier transform to decompose each image into frequency domains, enabling frequency-dependent fusion of features. Event information is fused into the high-frequency regions to enhance structural detail, while low-frequency image components are preserved to maintain color consistency. The framework builds upon an existing frequency-aware fusion model by introducing a Fourier-based regional feature selection mechanism to guide detail enhancement. Experimental evaluation on a public dataset demonstrates measurable improvements in image quality, as reflected by increases in peak signal-to-noise ratio and structural similarity index. Visual comparisons further confirm improved edge sharpness, richer textures, and more natural brightness transitions under extreme illumination conditions.