A mamba low-illumination image enhancement method combining light color decoupling and cross-guided space-frequency attention and a computer readable storage medium

Abstract

This application provides a Mamba low-light image enhancement method and computer-readable storage medium that combines light-color decoupling and cross-guided frequency-space attention, relating to the fields of digital image processing and computer vision. The method includes acquiring an input image and performing color space mapping; generating an illumination estimation input; obtaining an initial brightened image and illumination-perceived features through an illumination estimation module; extracting multi-level frequency domain features and spatial domain features; inputting the frequency domain features and spatial domain features at each level into a cross-domain enhancement network composed of multi-level frequency-space fusion units, fusing them step-by-step under the guidance of illumination-perceived features, and outputting the final enhanced image. Under the guidance of illumination-perceived features, this application constructs a dual-branch collaborative enhancement framework in the frequency and spatial domains. The frequency branch restores texture and high-frequency details, while the spatial branch restores edge structure and regional consistency. An illumination-perceived cross-domain fusion module is designed to improve the adaptability of the enhancement results to complex low-light scenes.

Images