Abstract:In this paper, we propose a high-quality sonar image generation model based on generative adversarial networks, which is used to generate user-controlled high-resolution and high-quality sonar images, so that the targets in these images have more obvious features and can be further used in engineering applications such as target detection and image classification. As a result of the development of underwater detection technology, sonar target detection technology is widely used in various engineering fields. High resolution and high-quality underwater sonar images have become an indispensable part of underwater detection. However, due to the high cost of sonar and the high cost of marine experiments, sonar datasets are scarce. Even then, few sonar data can be obtained, and most images have no target and have more noise so that they cannot be used for underwater detection. In recent years, using generative adversarial networks to generate images has gradually become an important method of solving the problem of data scarcity. But there are still some shortcomings — the images produced are poor and the resolution is low. To overcome the issue of poor image quality and low resolution caused by gradient disappearing and sudden change in the training process, we, therefore, use a controllable multi-layer transposed convolutional layer and gradient correction term to improve image resolution and image quality. A simple network structure does not make the model difficult to train, and the user may change the size of the generated image by simply changing the convolution parameter i to achieve the objective of improving the resolution. Moreover, the gradient correction term added in the training phase will always restrict the gradient to a certain value. To a certain degree, the problems of gradient disappearing and gradient sudden change in model training can be overcome, thereby providing a guarantee for the production of high-quality images. Experiments demonstrate that this method will successfully increase the resolution and image quality of the GAN-based image and make the image target more apparent.

S^2RGAN : Sonar-Image Super-Resolution Based on Generative Adversarial Network

Deeper super-resolution generative adversarial network with gradient penalty for sonar image enhancement

A Multi-Scale Generative Adversarial Network for Underwater Sonar Image Super-Resolution

Simultaneous restoration and super-resolution GAN for underwater image enhancement

Super-Resolution Reconstruction of Underwater Image Based on Image Sequence Generative Adversarial Network

Recovering Super-Resolution Generative Adversarial Network For Underwater Images

High-Quality Sonar Image Generation Algorithm Based on Generative Adversarial Networks

DAE-GAN: Underwater Image Super-Resolution Based on Symmetric Degradation Attention Enhanced Generative Adversarial Network

Learning Spectral and Spatial Features Based on Generative Adversarial Network for Hyperspectral Image Super-Resolution

Hyperspectral Imagery Spatial Super-Resolution Using Generative Adversarial Network.

End-to-End Image Super-Resolution Via Generative Adversarial Network

Learning Structral Coherence Via Generative Adversarial Network for Single Image Super-Resolution

Single Remote Sensing Image Super-Resolution Via a Generative Adversarial Network with Stratified Dense Sampling and Chain Training

Generative Adversarial Network-Based Super-Resolution Considering Quantitative And Perceptual Quality

Underwater Sonar Image Classification Using Generative Adversarial Network and Convolutional Neural Network

SD-GAN: Saliency-Discriminated GAN for Remote Sensing Image Superresolution

Single Image Super-Resolution Based on Enhanced Deep Residual GAN

Improved Residual Dense Network for Large Scale Super-Resolution via Generative Adversarial Network

SRPGAN: Perceptual Generative Adversarial Network for Single Image Super Resolution.

SHBGAN: Hybrid Bilateral Attention GAN for Seismic Image Super-Resolution Reconstruction

Underwater Image Enhancement Via Efficient Generative Adversarial Network