Underwater Image Co-Enhancement With Correlation Feature Matching and Joint Learning

Qi, Qi; Zhang, Yongchang; Tian, Fei; Wu, Qi; Li, Kunqian; Luan, Xin; Song, Dalei

doi:10.1109/tcsvt.2021.3074197

Cited by 72 publications

(23 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In The traditional approaches are the dark channel prior (DCP) method [52], the gamma correction (GC) method, the UCM method [25], the Retinex-based method [26], the histogram prior-based method (HP) [20], LAB-MSR [27], the two-step method [16] and the hybrid method [17]. The deep learningbased competitors include two recent GAN-based approaches (MLFcGAN [46] and FUnIE-GAN [47]) and four deep models trained with paired training samples (the U-Net-based enhancement model [53], WaterNet [7], Ucolor [9] and UICoE-Net [8]).…”

Section: Comparison Methodsmentioning

confidence: 99%

“…In UIE-DAL [40], the enhancement network is also designed with an encoder-decoder structure, in which the encoder is trained with adversarial loss to produce water type-agnostic features. More recently, Qi et al [8] proposed a two-branch encoder-decoder co-enhancement network. Li et al [9] designed a Ucolor network with multi-color space encoder and a medium transmission-guided decoder.…”

Section: Encoder-decoder Structure-based Networkmentioning

confidence: 99%

“…Comparisons of underwater image enhancement results on a greenish underwater image from UIEB [7]. The enhancements of DCP [52], GC, UCM [25], Retinex [26], HP [20], LAB-MSR [27], Two-Step [16], Hybrid [17], MLFcGAN [46], FUnIE-GAN [47], U-Net [53], WaterNet [7], UICoE-Net [8], Ucolor [9], our methods with two different template-selection modules and the reference are presented for comparison A noteworthy phenomenon is that traditional non-learning methods usually perform much better than deep learning-based methods according to UIQM, UCIQE, e and r metrics, which is diametrically opposed to quantitative results of reference-based metrics. As discussed in [7], UIQM and UCIQE did not take color shift and artifacts into account.…”

Section: Figurementioning

confidence: 99%

“…In an underwater image enhancement dataset that provides both raw underwater images and corresponding clear references, such as the Underwater Image Enhancement Benchmark Enhancements with three latest deep enhancement models, that is, WaterNet [7], UICoE-Net [8], and Ucolor [9], which directly learn to model the mappings between degradation and enhancement. (II) Enhancement with the proposed latent consistency learning-based color transfer (UIEB) [7] dataset, we find that clear underwater images can provided a wealth of reference and constraint information for the color correction and contrast improvement of other raw underwater images.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Underwater image enhancement with latent consistency learning‐based color transfer

Yang

Tian

et al. 2022

IET Image Processing

Self Cite

View full text Add to dashboard Cite

Due to the inevitable wavelength-dependent light absorption and forward/backward scattering, underwater images usually suffer severe color distortion and are hazy. It has become quite necessary to improve the visual quality of underwater images for both underwater observation and operation. Traditional enhancement methods and existing deep learningbased approaches to underwater image enhancement usually produce unsatisfactory results for photographs taken in complicated, wild underwater scenes. In such scenes, complex and diverse degradation-enhancement mappings are often difficult to model, especially since there are very limited samples available for learning. Inspired by the success of colortransfer techniques, it is found that clear template image-assisted color transfer is a promising strategy for underwater image enhancement, including not only color correction but also contrast and visibility improvement. Therefore, instead of directly learning the complex deep enhancement models, it is proposed to select proper color-transfer templates by learning the latent consistency between the templates and the raw underwater images. The proposed new enhancement strategy alleviates the problem caused by incomplete colorcorrection models and provides more stable enhancements by utilizing color transfer with consideration of global color distribution consistency and local visual contrast. Comprehensive experiments conducted on UIEB, RUIE, URPC and SQUID datasets demonstrate the good performance and great potential of the proposed new underwater image enhancement strategy.

show abstract

Section: Comparison Methodsmentioning

confidence: 99%

Section: Encoder-decoder Structure-based Networkmentioning

confidence: 99%

Section: Figurementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Underwater image enhancement with latent consistency learning‐based color transfer

Yang

Tian

et al. 2022

IET Image Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, researches on obtaining depth information directly from images without using depth sensors have attracted increased attention. In recent years, deep learning has gained traction, and more deep learning models have been developed for image processing and image quality recovery (e.g., depth estimation, air pollution detection [1], image enhancement [2]). The superior results of efficient network structures such as fully convolutional neural networks and ResNet [3] have indicated that neural networks are superior to hand-crafted features.…”

Section: Introductionmentioning

confidence: 99%

Semantically guided self‐supervised monocular depth estimation

Sun

Wang

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Depth information plays an important role in the vision‐related activities of robots and autonomous vehicles. An effective method to obtain 3D scene information is self‐supervised monocular depth estimation, which utilizes large and diverse monocular video datasets during the training process without the need for ground‐truth data. A novel multi‐task learning strategy that uses semantic information to guide the monocular depth estimation method while maintaining self‐supervision is proposed. An improved differential direct visual odometer (DDVO) combined with Pose‐Net is applied for achieving better pose prediction. Minimum reprojection loss with auto‐masking and semantic masking is used to remove the effects of low‐texture areas and moving dynamic‐class objects within scenes. Concurrently, the semantic masking is introduced into the DDVO pose predictor to filter moving objects and reduce the matching error between monocular sequence frames. In addition, PackNet is employed as the backbone of multi‐task learning to further improve the accuracy of deep prediction. The proposed method produces state‐of‐the‐art results for monocular depth estimation on the KITTI Eigen split benchmark, even outperforming supervised methods that have been trained using ground‐truth depth.

show abstract

Uncertainty Inspired Underwater Image Enhancement

Huang

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Underwater Image Co-Enhancement With Correlation Feature Matching and Joint Learning

Cited by 72 publications

References 55 publications

Underwater image enhancement with latent consistency learning‐based color transfer

Underwater image enhancement with latent consistency learning‐based color transfer

Semantically guided self‐supervised monocular depth estimation

Uncertainty Inspired Underwater Image Enhancement

Contact Info

Product

Resources

About