This repository contains our python implementation of the IJCV 2023 paper, Improving Visual Perception Via Adaptive Dark Pixel Prior and Color Correction. If you use any code or data from our work, please cite our paper.

@article{zhou2023underwater, title={Underwater camera: improving visual perception via adaptive dark pixel prior and color correction}, author={Zhou, Jingchun and Liu, Qian and Jiang, Qiuping and Ren, Wenqi and Lam, Kin-Man and Zhang, Weishi}, journal={International Journal of Computer Vision}, pages={1--19}, year={2023}, publisher={Springer} }

(1). We provide scripts that make it easier to test data. The following are the steps:
(2). Download code and comile.
	You need to install the dependencies in require.txt.
(3). Download dataset to "InputImages" folder.
(4). Change "file=." to corresponding path.
(5). Run project.

• If you get an error using the eng.niqe() function, configure the Python call matlab environment and create the niqe.m file that uses the matlab built-in function niqe().
• If the output is incorrect, try using the colorCheck() function.

You can find results in folder "OutputImages".

Baidu Cloud Disk： Link: https://pan.baidu.com/s/1cgHGNRAz7Fs8lbwckuWZGA Access code: 8d2p

We present a novel method for underwater image restoration, which combines a Comprehensive Imaging Formation Model with prior knowledge and unsupervised techniques. Our approach has two main components: depth map estimation using a Channel Intensity Prior (CIP) and backscatter elimination through Adaptive Dark Pixels (ADP). The CIP effectively mitigates issues caused by solid-colored objects and highlighted regions in underwater scenarios. The ADP, utilizing a dynamic depth conversion, addresses issues associated with narrow depth ranges and backscatter. Furthermore, an unsupervised method is employed to enhance the accuracy of monocular depth estimation and reduce artificial illumination influence. The final output is refined via color compensation and a blue-green channel color balance factor, delivering artifact-free images. Experimental results show that our approach outperforms state-of-the-art methods, demonstrating its efficacy in dealing with uneven lighting and diverse underwater environments.