When Schrödinger Bridge Meets Real-World Image Dehazing with Unpaired Training
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909687092346880 |
|---|---|
| author | Lan, Yunwei Cui, Zhigao Luo, Xin Liu, Chang Wang, Nian Zhang, Menglin Su, Yanzhao Liu, Dong |
| author_facet | Lan, Yunwei Cui, Zhigao Luo, Xin Liu, Chang Wang, Nian Zhang, Menglin Su, Yanzhao Liu, Dong |
| contents | Recent advancements in unpaired dehazing, particularly those using GANs, show promising performance in processing real-world hazy images. However, these methods tend to face limitations due to the generator's limited transport mapping capability, which hinders the full exploitation of their effectiveness in unpaired training paradigms. To address these challenges, we propose DehazeSB, a novel unpaired dehazing framework based on the Schrödinger Bridge. By leveraging optimal transport (OT) theory, DehazeSB directly bridges the distributions between hazy and clear images. This enables optimal transport mappings from hazy to clear images in fewer steps, thereby generating high-quality results. To ensure the consistency of structural information and details in the restored images, we introduce detail-preserving regularization, which enforces pixel-level alignment between hazy inputs and dehazed outputs. Furthermore, we propose a novel prompt learning to leverage pre-trained CLIP models in distinguishing hazy images and clear ones, by learning a haze-aware vision-language alignment. Extensive experiments on multiple real-world datasets demonstrate our method's superiority. Code: https://github.com/ywxjm/DehazeSB. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_09524 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | When Schrödinger Bridge Meets Real-World Image Dehazing with Unpaired Training Lan, Yunwei Cui, Zhigao Luo, Xin Liu, Chang Wang, Nian Zhang, Menglin Su, Yanzhao Liu, Dong Computer Vision and Pattern Recognition Recent advancements in unpaired dehazing, particularly those using GANs, show promising performance in processing real-world hazy images. However, these methods tend to face limitations due to the generator's limited transport mapping capability, which hinders the full exploitation of their effectiveness in unpaired training paradigms. To address these challenges, we propose DehazeSB, a novel unpaired dehazing framework based on the Schrödinger Bridge. By leveraging optimal transport (OT) theory, DehazeSB directly bridges the distributions between hazy and clear images. This enables optimal transport mappings from hazy to clear images in fewer steps, thereby generating high-quality results. To ensure the consistency of structural information and details in the restored images, we introduce detail-preserving regularization, which enforces pixel-level alignment between hazy inputs and dehazed outputs. Furthermore, we propose a novel prompt learning to leverage pre-trained CLIP models in distinguishing hazy images and clear ones, by learning a haze-aware vision-language alignment. Extensive experiments on multiple real-world datasets demonstrate our method's superiority. Code: https://github.com/ywxjm/DehazeSB. |
| title | When Schrödinger Bridge Meets Real-World Image Dehazing with Unpaired Training |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2507.09524 |