# Driving-Video Dehazing with Non-Aligned Regularization for Safety Assistance

Source: [Driving-Video Dehazing with Non-Aligned Regularization for Safety Assistance](<https://www.sota2.com/research/paper/68fa775b1e209131cea72ab5>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T04:49:04.596Z

Authors: Junkai Fan, Jiangwei Weng, Kun Wang, Yijun Yang, Jianjun Qian, Jun Li, Jian Yang

Published: 2024-05-16T11:28:01.000Z

## Abstract

Real driving-video dehazing poses a significant challenge due to the inherent difficulty in acquiring precisely aligned hazy/clear video pairs for effective model training, especially in dynamic driving scenarios with unpredictable weather conditions. In this paper, we propose a pioneering approach that addresses this challenge through a nonaligned regularization strategy. Our core concept involves identifying clear frames that closely match hazy frames, serving as references to supervise a video dehazing network. Our approach comprises two key components: reference matching and video dehazing. Firstly, we introduce a non-aligned reference frame matching module, leveraging an adaptive sliding window to match high-quality reference frames from clear videos. Video dehazing incorporates flow-guided cosine attention sampler and deformable cosine attention fusion modules to enhance spatial multiframe alignment and fuse their improved information. To validate our approach, we collect a GoProHazy dataset captured effortlessly with GoPro cameras in diverse rural and urban road environments. Extensive experiments demonstrate the superiority of the proposed method over current state-of-the-art methods in the challenging task of real driving-video dehazing. Project page.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Dehazing on DrivingHazy NoRef (test)](<https://www.sota2.com/research/sota/dehazing-on-drivinghazy-noref-test.md>): Dehazing; DrivingHazy NoRef (test); FADE: 82.07.

- [Dehazing on InternetHazy (test)](<https://www.sota2.com/research/sota/dehazing-on-internethazy-test.md>): Dehazing; InternetHazy (test); FADE: 0.8745.

- [Video Dehazing on REVIDE](<https://www.sota2.com/research/sota/video-dehazing-on-revide.md>): Video Dehazing; REVIDE; PSNR: 24.34.

- [Object Detection on UHV-4K](<https://www.sota2.com/research/sota/object-detection-on-uhv-4k.md>): Object Detection; UHV-4K; Detection Rate: 14.7.

- [Semantic Segmentation on UHV-4K](<https://www.sota2.com/research/sota/semantic-segmentation-on-uhv-4k.md>): Semantic segmentation; UHV-4K; Boundary Error (BdryE): 0.0079.

- [Image Quality Assessment on UHV-4K](<https://www.sota2.com/research/sota/image-quality-assessment-on-uhv-4k.md>): Image Quality Assessment; UHV-4K; NIQE: 10.43.

- [Video Dehazing on UHV-4K (test)](<https://www.sota2.com/research/sota/video-dehazing-on-uhv-4k-test.md>): Video Dehazing; UHV-4K (test); tOF: 0.004551.

- [Video Dehazing on UHV-4K](<https://www.sota2.com/research/sota/video-dehazing-on-uhv-4k.md>): Video Dehazing; UHV-4K; PSNR: 21.02.

- [Video Dehazing on HazeWorld](<https://www.sota2.com/research/sota/video-dehazing-on-hazeworld.md>): Video Dehazing; HazeWorld; PSNR: 27.05.

- [Dehazing on GoProHazy (test)](<https://www.sota2.com/research/sota/dehazing-on-goprohazy-test.md>): Dehazing; GoProHazy (test); FADE: 0.7598.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2405.09996>)

[PDF](<https://arxiv.org/pdf/2405.09996.pdf>)