Zero-shot Single Image Restoration through Controlled Perturbation of Koschmieder’s Model

Kar, Aupendu; Dhara, Sobhan Kanti; Sen, Debashis; Biswas, Prabir Kumar

doi:10.1109/cvpr46437.2021.01594

Cited by 40 publications

(46 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The work (Li et al 2021b) proposed an underwater image restoration network via medium transmission guided multi-color space embedding. Currently, a few "zero-shot" approaches (Kar et al 2021;Gandelsman, Shocher, and Irani 2019) have been proposed, which perform the restoration using a single image.…”

Section: Learning-based Methodsmentioning

confidence: 99%

“…Recently, "zero-shot" approaches (Li et al 2021a;Kar et al 2021;Gandelsman, Shocher, and Irani 2019;Shocher, Cohen, and Irani 2018) have been developed which learn the restoration task using a single image and train a small image-specific network at test time. As such "zero-shot" methods do not use other supervision information beyond the input image itself, they require to leverage self/unsupervised losses or regularizers for training.…”

Section: Introductionmentioning

confidence: 99%

“…As such "zero-shot" methods do not use other supervision information beyond the input image itself, they require to leverage self/unsupervised losses or regularizers for training. For example, the unsupervised single image restoration (Kar et al 2021) estimates the parameters of Koschmieder's model through a controlled perturbation. The unsupervised image decomposition method (Gandelsman, Shocher, and Irani 2019) applies multiple "Deep-image-Prior" (DIP) networks (Ulyanov, Vedaldi, and Lempitsky 2018) and the power of the internal patch recurrence to separate images into their basic components.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Unsupervised Underwater Image Restoration: From a Homology Perspective

Lin

Yan

et al. 2022

AAAI

View full text Add to dashboard Cite

Underwater images suffer from degradation due to light scattering and absorption. It remains challenging to restore such degraded images using deep neural networks since real-world paired data is scarcely available while synthetic paired data cannot approximate real-world data perfectly. In this paper, we propose an UnSupervised Underwater Image Restoration method (USUIR) by leveraging the homology property between a raw underwater image and a re-degraded image. Specifically, USUIR first estimates three latent components of the raw underwater image, i.e., the global background light, the transmission map, and the scene radiance (the clean image). Then, a re-degraded image is generated by randomly mixing up the estimated scene radiance and the raw underwater image. We demonstrate that imposing a homology constraint between the raw underwater image and the re-degraded image is equivalent to minimizing the restoration error and hence can be used for the unsupervised restoration. Extensive experiments show that USUIR achieves promising performance in both inference time and restoration quality.

show abstract

Section: Learning-based Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Underwater Image Restoration: From a Homology Perspective

Lin

Yan

et al. 2022

AAAI

View full text Add to dashboard Cite

show abstract

“…Berman et al [10] recovered the underwater image color by estimating blue-red and blue-green channel attenuation ratios for different types of underwater images, and collected a new underwater dataset with real reference images. Kar et al [11] proposed a real image restoration model based on the image degradation characteristics derived from the Koschmieder model, which showed good results on multiple real image datasets.…”

Section: Restoration-based Methodsmentioning

confidence: 99%

Attention-based for Multiscale Fusion Underwater Image Enhancement

2022

KSII TIIS

View full text Add to dashboard Cite

Underwater images often suffer from color distortion, blurring and low contrast, which is caused by the propagation of light in the underwater environment being affected by the two processes: absorption and scattering. To cope with the poor quality of underwater images, this paper proposes a multiscale fusion underwater image enhancement method based on channel attention mechanism and local binary pattern (LBP). The network consists of three modules: feature aggregation, image reconstruction and LBP enhancement. The feature aggregation module aggregates feature information at different scales of the image, and the image reconstruction module restores the output features to high-quality underwater images. The network also introduces channel attention mechanism to make the network pay more attention to the channels containing important information. The detail information is protected by real-time superposition with feature information. Experimental results demonstrate that the method in this paper produces results with correct colors and complete details, and outperforms existing methods in quantitative metrics.

show abstract

“…We compare our methods against several state-of-the-art LLIE baselines. The baselines can be classified into three categories, the supervised methods (Retinex-Net [10] and KinD [13]), the unsupervised methods (EnlightenGAN [11] and Zero-DCE [12]), and the zero-shot methods (LIME [9] and Kar et al [123]). All the baselines are implemented using the publicly available codes as well as recommended parameters.…”

Section: Experiments Settingmentioning

confidence: 99%

Deep reinforcement learning for intractable routing & inverse problems

Zhang¹

View full text Add to dashboard Cite

show abstract

Zero-shot Single Image Restoration through Controlled Perturbation of Koschmieder’s Model

Cited by 40 publications

References 56 publications

Unsupervised Underwater Image Restoration: From a Homology Perspective

Unsupervised Underwater Image Restoration: From a Homology Perspective

Attention-based for Multiscale Fusion Underwater Image Enhancement

Deep reinforcement learning for intractable routing & inverse problems

Contact Info

Product

Resources

About