Underwater ghost imaging based on generative adversarial networks with high imaging quality.

In recent years, enhancement of underwater images based on generative adversarial network (GAN) has been widely used. Aiming at the shortcoming that the texture details of the images generated by GAN are not clear enough, an underwater image pre-process unit and a generator including a perceptual subnet and a refine subnet are designed to improve visual quality of underwater images. The role of the perceptual subnet is to maintain the structure and semantic information of the input image. Refine subnet is to make the generated image clearer in the details of the texture. Specially, considering that underwater images may prevent the network from learning effectively, an underwater image pre-process unit is proposed to improve the quality of underwater images. To maintain the structure and semantic information of underwater images and preserve texture details of underwater images, the perceptual subnet and the refine subnet are proposed, respectively. Then, color-structure perception loss is proposed to obtain good performance in color and structure, and content loss and detail loss are proposed to keep content consistent and sharper texture. Ablation study verify the rationality and effectiveness of each loss function. Finally, subjective and objective experiments prove that the proposed method can obtain underwater images with higher visual quality and clearer texture details.

Underwater images often exhibit detail blurring and color distortion due to light scattering, impurities, and other influences, obscuring essential textures and details. This presents a challenge for existing super-resolution techniques in identifying and extracting effective features, making high-quality reconstruction difficult. This research aims to innovate underwater image super-resolution technology to tackle this challenge. Initially, an underwater image degradation model was created by integrating random subsampling, Gaussian blur, mixed noise, and suspended particle simulation to generate a highly realistic synthetic dataset, thereby training the network to adapt to various degradation factors. Subsequently, to enhance the network’s capability to extract key features, improvements were made based on the symmetrically structured blind super-resolution generative adversarial network (BSRGAN) model architecture. An attention mechanism based on energy functions was introduced within the generator to assess the importance of each pixel, and a weighted fusion strategy of adversarial loss, reconstruction loss, and perceptual loss was utilized to improve the quality of image reconstruction. Experimental results demonstrated that the proposed method achieved significant improvements in peak signal-to-noise ratio (PSNR) and underwater image quality measure (UIQM) by 0.85 dB and 0.19, respectively, significantly enhancing the visual perception quality and indicating its feasibility in super-resolution applications.

Image dehazing has always been a challenging topic in image processing. The development of deep learning methods, especially the generative adversarial networks (GAN), provides a new way for image dehazing. In recent years, many deep learning methods based on GAN have been applied to image dehazing. However, GAN has two problems in image dehazing. Firstly, For haze image, haze not only reduces the quality of the image but also blurs the details of the image. For GAN network, it is difficult for the generator to restore the details of the whole image while removing the haze. Secondly, GAN model is defined as a minimax problem, which weakens the loss function. It is difficult to distinguish whether GAN is making progress in the training process. Therefore, we propose a guided generative adversarial dehazing network (GGADN). Different from other generation adversarial networks, GGADN adds a guided module on the generator. The guided module verifies the network of each layer of the generator. At the same time, the details of the map generated by each layer are strengthened. Network training is based on the pre-trained VGG feature model and L1-regularized gradient prior which is developed by new loss function parameters. From the dehazing results of synthetic images and real images, the proposed method is better than the state-of-the-art dehazing methods.

Magnetic Particle Imaging (MPI), as an emerging non-invasive imaging technique, acquires high-resolution images by detecting the response of magnetic nanoparticles. Reconstruction based on System Matrix (SM) is an important component of MPI research. However, when scanning parameters, particle types, or environmental conditions change, SM requires repetitive and time-consuming calibration scans, significantly reducing its practicality. In this study, we propose a novel SM super-resolution recovery method, PGSM-net, based on Generative Adversarial Network (GAN) and Pyramid Attention Mechanism (PAM) to enhance SM calibration in MPI. The generator utilizes enhanced ResNet blocks and a PAM module for multi-scale feature extraction, attention weighting, and feature fusion to recover finer details from low-resolution inputs. The discriminator, an unpooled VGG network, assesses high-dimensional feature consistency to aid the generator in producing more realistic and higher-quality images. Our approach represents the first application of GAN and perceptual loss in MPI system matrix calibration. Experimental results show significant improvements in SM recovery quality and detail, highlighting its potential in MPI applications.

Underwater imaging has long been focused on dehazing and color correction to address severe degradation in the water medium. In this paper, we propose a learning-based image restoration method that uses Generative Adversarial Networks (GAN). For network generality and learning flexibility, we constituted unpaired image translation frameworks into image restoration. The proposed method utilizes multiple cyclic consistency losses that capture image characteristics and details of underwater images. To prepare unpaired images of clean and degraded scenes, we collected images from Flickr and filtered out false images using image characteristics. For validation, we extensively evaluated the proposed network on simulated and real underwater hazy images. Also, we tested our method on conventional computer vision algorithms, such as the level of edges and feature matching results.

Traditional deblurring algorithms such as blur kernel estimation and gradient prior methods, these methods can only deal with a single type of blurry image, and such methods are very time-consuming. Due to the excellent performance of deep learning neural networks in the field of image generation, this paper regards the traditional image deblurring problem as a special image transformation problem. First, a multi-scale network is used to restore the blurry image, the network can restore the image in an end-to-end manner. Then, the multi-scale network will output a restored image. Finally, the generative adversarial network (GAN) can make the restored image closer to the real clear image in the sense. In terms of loss function, perceptual loss in content loss is adopted to ensure the consistency of the generated image and clear image in content. The results of deblurring will be evaluated by testing the peak signal to noise ratio (PSNR) and structural similarity (SSIM) between the generated image and the clear image. The results show that the PSNR and SSIM of the algorithm in the GoPro dataset have reached 28.01 and 0.8538, respectively, and can recover some details lost due to shake blur.

Fast transformation of discriminators into encoders using pre-trained GANs

Underwater image enhancement has always been a hot spot in underwater vision research. However, due to complicated underwater environment, a lot of problems such as the color distortion and low brightness of underwater raw images are very likely to occur. In response to the above situation, we proposed a generative adversarial network that integrated multiple attention to enhance underwater images. In the generator, we introduced multi-layer dense connections and CSAM modules, of which the former could capture more detailed features and make use of previous features, while the latter could improve the utilization of the feature map. Meanwhile, we improved the enhancement effect of the generated image by combining VGG19 content loss function and SmoothL1 loss function. Finally, we verified the effectiveness of the proposed model through qualitative and quantitative experiments, and compared the results with the performance of several latest models. The results show that the methods proposed in this paper are superior to the existing methods.

Images captured by underwater robots often suffer from issues such as blurring and colour distortion, which hinder effective feature extraction and target recognition in underwater environments. To address these challenges, this paper proposes a novel underwater image enhancement method based on generative adversarial networks (GANs), termed multiple colour space underwater generative adversarial network (MCS‐UGAN). The proposed method is built upon a GAN framework, consisting of a generator and a discriminator. The generator comprises two main modules: a deblurring module and a colour correction module. The deblurring module innovatively incorporates an efficient multi‐scale feature extraction technique and an attention mechanism, which enhances object contours while preserving fine image details. The colour correction module integrates residual blocks into the U‐Net architecture, effectively mitigating the problems of gradient vanishing and explosion during backpropagation in underwater image enhancement networks, thereby enhancing the network's feature learning capability. This design corrects colour distortions while preserving edge information in the image. The discriminator adopts the PatchGAN structure, which focuses on the local regions of the image, significantly improving the generator's ability to restore high‐frequency details and thus enhancing the quality of the generated images. Experimental results on benchmark datasets demonstrate that, compared to existing methods, MCS‐UGAN achieves superior performance in terms of peak signal‐to‐noise ratio, structural similarity index measure, underwater image quality measure, and underwater colour image quality evaluation, with average values of 26.24, 0.91, 3.13, and 0.64, respectively. Results from real‐world applications further show that MCS‐UGAN effectively increases the number of extracted corner points, validating its practicality and effectiveness. The code is available at https://github.com/invincibility6/MCS‐UGAN.git

This study presents a generative adversarial network (GAN) called conditional Wasserstein GAN‐gradient penalty (CWGAN‐GP)&DenseNet and ResNet, and a convolutional neural network (CNN) called improved CNN to complete underwater sonar image classification. Specifically, to solve the problem of insufficient underwater sonar image data, the CWGAN‐GP&DR is developed to expand underwater sonar image data set. Besides, to improve the analysis and utilisation of the feature map and reduce the misclassification rate of categories with similar probabilities, improved CNN is proposed to complete the final underwater sonar image classification. Finally, compared with other methods, the CWGAN‐GP&DR generate better underwater sonar images and effectively expand the underwater sonar image data set. Moreover, compared with the original data set and other expanded data set, the highest accuracy rate of 85.00% can be obtained on the CWGAN‐GP&DR expanded data set by CNN. Furthermore, CNN, CNN‐bais and improved CNN are used to perform classification experiments on each data set, and the accuracy of the improved CNN is the highest on all data sets and reached the highest accuracy of 87.71% on CWGAN‐GP&DR expanded data set. The experimental results demonstrate that the proposed method can effectively improve the performance of underwater sonar image classification.

Ghost imaging of the low or high frequency based on the corresponding spatial-frequency of the reference pattern

In practical thermal light ghost imaging, the nonuniform intensity distribution of the thermal light applied causes distortion of the retrieved image. We propose uniformly weighted arithmetics, including uniformly weighted ghost imaging and uniformly weighted differential ghost imaging (UWDGI), to improve the imaging quality of ghost imaging with nonuniform thermal light fields. Two kinds of nonuniform thermal light fields are discussed, including planar Gaussian distributed thermal light and near-field three-dimensional thermal light. The retrieved image using a uniformly weighted arithmetic has a lower distortion than that using the corresponding traditional ghost imaging arithmetic. In addition, the best imaging quality is achieved with the UWDGI arithmetic.

High-quality direct ghost imaging of random dynamic targets based on convolutional neural network

Underwater image enhancement based on conditional generative adversarial network

Ghost imaging with single-pixel bucket detector has attracted more and more current attention due to its marked physical characteristics. However, in ghost imaging, a large number of reference intensity patterns are usually required for object reconstruction, hence many applications based on ghost imaging (such as tomography and optical security) may be tedious since heavy storage or transmission is requested. In this paper, we report that the compressed reference intensity patterns can be used for object recovery in computational ghost imaging (with single-pixel bucket detector), and object verification can be further conducted. Only a small portion (such as 2.0% pixels) of each reference intensity pattern is used for object reconstruction, and the recovered object is verified by using nonlinear correlation algorithm. Since statistical characteristic and speckle averaging property are inherent in ghost imaging, sidelobes or multiple peaks can be effectively suppressed or eliminated in the nonlinear correlation outputs when random pixel positions are selected from each reference intensity pattern. Since pixel positions can be randomly selected from each 2D reference intensity pattern (such as total measurements of 20000), a large key space and high flexibility can be generated when the proposed method is applied for authenticationbased cryptography. When compressive sensing is used to recover the object with a small number of measurements, the proposed strategy could still be feasible through further compressing the recorded data (i.e., reference intensity patterns) followed by object verification. It is expected that the proposed method not only compresses the recorded data and facilitates the storage or transmission, but also can build up novel capability (i.e., classical or quantum information verification) for ghost imaging.