The paper presents NightHaze, a novel method for nighttime image dehazing using self-prior learning with severe augmentation. By intentionally degrading clear images with light effects and noise, the model learns robust priors. The proposed approach outperforms existing methods significantly in terms of performance metrics.
The study addresses the challenges of nighttime image dehazing by introducing a unique method that leverages severe augmentation to improve visibility in hazy images. The self-prior learning technique effectively enhances the clarity of nighttime scenes by suppressing glow and revealing details. Additionally, a self-refinement module is proposed to address artifacts caused by over-suppression.
Extensive experiments demonstrate the effectiveness of NightHaze in improving visibility in real-world nighttime haze images. The method achieves state-of-the-art performance, surpassing existing techniques by a substantial margin. Overall, NightHaze shows promise in enhancing visibility and clarity in challenging nighttime conditions.
To Another Language
from source content
arxiv.org
ข้อมูลเชิงลึกที่สำคัญจาก
by Beibei Lin,Y... ที่ arxiv.org 03-13-2024
https://arxiv.org/pdf/2403.07408.pdfสอบถามเพิ่มเติม