UIE_survey

In fact, we don't know what we're doing. A project that is trying to collaborate on underwater image enhancement. We want to classify the papers according to some rule We hope every single article has an abstract in one sentence, including function, database, GT, appearance and so on. Articles shall be in recent 3 years to keep up with the advanced feature.

supervised

semi-supervised

(2023.10) Dimma: Semi-supervised Low Light Image Enhancement with Adaptive Dimming

DOI: 10.48550/arXiv.2310.09633

Method:

Apperance:

Repository: https://github.com/WojciechKoz/Dimma

unsupervised

(2023.5) Real-World Underwater Image Enhancement Based on Attention U-Net

DOI: https://doi.org/10.3390/jmse11030662

Method: A conditional generative adversarial network model based on attention U-Net which contains an attention gate mechanism

Database: UIEB, EUVP

Appearance:

(2023.4) Underwater Image Enhancement Based on Zero-Reference Deep Network

DOI: 10.1109/JOE.2023.3245686

Method:

NetWork:

Database: Laboratory(Unavailable), UIEB, RUIE

Apperance:

(2023.2) UMGAN: Underwater Image Enhancement Network for Unpaired Image-to-Image Translation

DOI: 10.3390/jmse11020447

Method: underwater multiscene generative adversarial network (UMGAN)

Dataset: UIEB, EUVP

Appearance:

(2022.2) Two-step domain adaptation for underwater image enhancement

DOI: 10.1016/j.patcog.2021.108324

Method:Transfer learning

(2021.12) Unpaired Underwater Image Enhancement Based on CycleGAN

DOI: 10.3390/info13010001

Method: Cycle generative adversarial network (UW-CycleGAN)

Dataset: URPC2019, EUVP

As shown in figure 2: The UW-CycleGAN architecture employs two generators and two discriminators to enhance underwater images. The generators are responsible for transforming the original images to enhanced images and back, ensuring content consistency; the discriminators evaluate the realism of the images. The process does not require paired data, relying instead on cycle consistency and content preservation to learn image enhancement.

(2021.1) EnlightenGAN: Deep Light Enhancement Without Paired Supervision

DOI: 10.1109/TIP.2021.3051462

Method:

Appearance:

(2019.7) Underwater Image Enhancement With a Deep Residual Framework

DOI: 10.1109/ACCESS.2019.2928976

Method:

Database:Public Database

(2019.6) Underwater Image Enhancement Using a Multiscale Dense Generative Adversarial Network

DOI: 10.1109/JOE.2019.2911447

Method:

Apperance:

Database: Laboratory(Unavailable)

others

(2019.11) An Underwater Image Enhancement Benchmark Dataset and Beyond

DOI: 10.1109/TIP.2019.2955241

Database: UIEB

(2020.1) Real-World Underwater Enhancement: Challenges, Benchmarks, and Solutions Under Natural Light

DOI: 10.1109/TCSVT.2019.2963772

Database: RUIE(UIQS, UCCS, UHTS)

(2022.1) Underwater Image Enhancement via Minimal Color Loss and Locally Adaptive Contrast Enhancement

DOI: 10.1109/TIP.2022.3177129

Method: Locally Adaptive Color Correction Method

Database: UCCS, UIQS, UIEB

Apperance:

(2019) Natural-based underwater image color enhancement through fusion of swarm-intelligence algorithm

DOI: https://doi.org/10.1016/j.asoc.2019.105810

Method: Natural-based underwater image color enhancement (NUCE)

Highlights • The proposed natural-based underwater image color enhancement enhances contrast and color. • Proposed NUCE method superimposes color cast neutralization, dual image fusion, and mean equalization steps. • The output images produce significant contrast and color while outstandingly addressing blue–green color cast. • Both qualitative and quantitative evaluations show better improvement of the addressed problems.

Database: UCCS, UIQS, UIEB

Function: