一种基于码本引导的夜间交通低光增强与过曝抑制方法
By constructing a codebook-guided image enhancement model, the problems of global exposure inconsistency and texture coupling in image enhancement methods in nighttime traffic scenes are solved. This model achieves brightness enhancement in dark areas and suppression of overexposure of vehicle lights and streetlights, thereby improving image quality and consistency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUNAN UNIV OF SCI & TECH
- Filing Date
- 2026-06-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing low-light image enhancement methods are prone to problems in nighttime traffic scenes, such as inconsistent global exposure, noise and texture coupling, overexposure diffusion of vehicle and street lights, color overflow, and loss of structural details around the light source.
A codebook-guided image enhancement model is constructed, including a codebook construction branch, a codebook-guided state space modeling branch, a wavelet frequency domain enhancement branch, an overexposed light source layer estimation branch, and a light source suppression-guided fusion module. Stable visual priors are provided through discrete visual codebooks. Combined with Mamba state space modeling and wavelet transform, the strong light core, halo diffusion, and transition region are explicitly estimated, and region-selective compensation and fusion are performed.
It improves visibility in dark areas, naturalness of light source areas, global exposure consistency, and texture detail retention, significantly enhancing the PSNR, SSIM, and OE-MAE performance of images.
Smart Images
Figure CN122415403A_ABST