Metalantis: A Comprehensive Underwater Image Enhancement Framework

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2024, Volume and Issue: 62, P. 1 - 19

Published: Jan. 1, 2024

Underwater images normally suffer from visual degradation issues such as color deviations, low contrasts, and blurred details. Recently, numerous underwater image enhancement algorithms have been proposed to address these issues. However, constrained by conditions, acquiring non-underwater depth maps for is often challenging. This limitation significantly hampers the performance of data driven-based methods physical model-based methods. Additionally, existing typically require manual parameter settings, which tend be bruteforce insufficient effectively diverse scenes. To overcome limitations, this paper presents a comprehensive framework comprising three phases: metamergence (i.e., meta submergence), metalief relief), metaebb ebb). These phases are dedicated virtual synthesis, map estimation, configuration state-of-the-art models reinforcement learning, separately. While trained separately, former phase provides necessary training latter. We refer overall metalantis Atlantis) because its processes, involving variations submergence via relief ebb over indoor scenes, mimic Atlantis. The empowers imaging through learning with virtually generated data. well-trained can take an sole input, process it into representations, finally enhance it. Comprehensive qualitative quantitative empirical evaluations validate that our outperforms release code at https://gitee.com/wanghaoupc/Metalantis_UIE.

Language: Английский

---

Self-organized underwater image enhancement

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2024, Volume and Issue: 215, P. 1 - 14

Published: June 29, 2024

---

INSPIRATION: A reinforcement learning-based human visual perception-driven image enhancement paradigm for underwater scenes

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108411 - 108411

Published: April 9, 2024

Language: Английский

---

A Pixel Distribution Remapping and Multi-Prior Retinex Variational Model for Underwater Image Enhancement

IEEE Transactions on Multimedia, Journal Year: 2024, Volume and Issue: 26, P. 7838 - 7849

Published: Jan. 1, 2024

High-quality underwater imaging is crucial for exploration. However, particle scattering and light absorption by seawater significantly degrade image clarity. To address these issues, we propose a novel enhancement (UIE) method that combines pixel distribution remapping (PDR) with multi-priority Retinex variational model. We design pre-compensation severely attenuated channels effectively prevents new color artifacts during correction. By combining the inter-channel coupling relationships, compute limiting factor to remap curves improve contrast. In addition, considering significant noise interference, introduce prior knowledge, including texture priors, while constructing model, penalty terms match characteristics remove excessive in reflectance component. Our approach efficiently decouples illumination components using rapid solver. Subsequently, gamma correction adjusts component, corrected are fused reconstruct final natural output image. Comprehensive evaluations across various datasets reveal our surpasses current state-of-the-art (SOTA) methods. These results demonstrate effectiveness of correcting bias compensating luminance losses imagery. code available at: https://github.com/zhoujingchun03/PDRMRV .

Language: Английский

---

Underwater image captioning: Challenges, models, and datasets

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2025, Volume and Issue: 220, P. 440 - 453

Published: Jan. 5, 2025

Language: Английский

---

HCMPE-Net: An unsupervised network for underwater image restoration with multi-parameter estimation based on homology constraint

Optics & Laser Technology, Journal Year: 2025, Volume and Issue: 186, P. 112616 - 112616

Published: Feb. 21, 2025

Language: Английский

---

GACA: A Gradient-Aware and Contrastive-Adaptive Learning Framework for Low-Light Image Enhancement

IEEE Transactions on Instrumentation and Measurement, Journal Year: 2024, Volume and Issue: 73, P. 1 - 14

Published: Jan. 1, 2024

Image gradients contain crucial information in the images. However, gradient of low-light images is often concealed darkness and susceptible to noise contamination. This imprecise poses a significant obstacle Low-Light Enhancement (LLIE) tasks. Simultaneously, methods relying solely on pixel-level reconstruction loss struggle accurately correct mapping from dimly lit normal images, resulting restored outcomes with color abnormalities or artifacts. In this paper, we propose Gradient-Aware Contrastive-Adaptive (GACA) Learning Framework address aforementioned issues. GACA initially estimates more accurate employs it as structural prior guide image generation. introduce novel regularization constraint better rectify mapping. Extensive experiments benchmark datasets downstream segmentation tasks demonstrate state-of-the-art performance generalization. Compared existing approaches, our method achieves an average 4.7% reduction NIQE datasets. The code available at https://github.com/iijjlk/GACA.

Language: Английский

---

Synergistic Multiscale Detail Refinement via Intrinsic Supervision for Underwater Image Enhancement

Proceedings of the AAAI Conference on Artificial Intelligence, Journal Year: 2024, Volume and Issue: 38(7), P. 7033 - 7041

Published: March 24, 2024

Visually restoring underwater scenes primarily involves mitigating interference from media. Existing methods ignore the inherent scale-related characteristics in scenes. Therefore, we present synergistic multi-scale detail refinement via intrinsic supervision (SMDR-IS) for enhancing scene details, which contain multi-stages. The low-degradation stage original images furnishes with achieved through feature propagation using Adaptive Selective Intrinsic Supervised Feature (ASISF) module. By supervision, ASISF module can precisely control and guide transmission across multi-degradation stages, minimizing irrelevant information stage. In encoder-decoder framework of SMDR-IS, introduce Bifocal Intrinsic-Context Attention Module (BICA). Based on principles, BICA efficiently exploits images. directs higher-resolution spaces by tapping into insights lower-resolution ones, underscoring pivotal role spatial contextual relationships image restoration. Throughout training, inclusion a loss function enhance network, allowing it to adeptly extract diverse scales. When benchmarked against state-of-the-art methods, SMDR-IS consistently showcases superior performance. Our code is available at https://github.com/zhoujingchun03/SMDR-IS

Language: Английский

---

Decoupled variational retinex for reconstruction and fusion of underwater shallow depth-of-field image with parallax and moving objects

Information Fusion, Journal Year: 2024, Volume and Issue: 111, P. 102494 - 102494

Published: May 25, 2024

Language: Английский

---

Dual branch Transformer-CNN parametric filtering network for underwater image enhancement

Journal of Visual Communication and Image Representation, Journal Year: 2024, Volume and Issue: 100, P. 104131 - 104131

Published: April 1, 2024

Language: Английский

---