$$\hbox {S}^2\hbox {RGAN}$$ S 2 RGAN : sonar-image super-resolution based on generative adversarial network
- PDF / 4,166,166 Bytes
- 15 Pages / 595.276 x 790.866 pts Page_size
- 40 Downloads / 209 Views
ORIGINAL ARTICLE
S2 RGAN: sonar-image super-resolution based on generative adversarial network Hongtao Song1 · Minghao Wang1 · Liguo Zhang1
· Yang Li1 · Zhengyi Jiang1 · Guisheng Yin1
Accepted: 2 October 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract As an important display mode of underwater environments, the sonar image has limitations on the resolution, which often leads to problems with low resolution of underwater objects. Therefore, the image super-resolution algorithm is needed to transform the images from low-resolution to high-resolution. It can improve the visual effect and contribute to subsequent processing such as 3D reconstruction and object recognition. This paper proposes a method for sonar image super-resolution based on generative adversarial networks (GAN). By comparing the super-resolution effects of various interpolation and convolutional neural network algorithms on sonar images, a Residual-in-Residual Dense Block network is employed as the generator of GAN since it has the low distortion and high perceptual quality. Because the sonar image training set does not have enough data, the generator utilizes the transfer learning on the sonar images to produce an optimized network model which is more suitable for super-resolution of sonar image. The VGG19 network is employed as the discriminator. In addition, the perceptual loss is introduced into the loss function of S2 RGAN to further improve the perceptual quality of super-resolution images. The experimental results indicate that the proposed S2 RGAN shows excellent performance. The generated super-resolution images of S2 RGAN have the remarkable advantages of both lower distortion and higher perceptual quality comparing with other methods. Because S2 RGAN focuses more on the reality and overall visual effect of super-resolution sonar images, it is suitable for various underwater situations. Keywords Sonar-image · Super-resolution · Generative adversarial network · Transfer learning
1 Introduction Until now, the sound wave has been the major transmission medium for humans to detect underwater objects. Sonar has the advantage of high cost performance and easy oper-
B
Liguo Zhang [email protected] Hongtao Song [email protected] Minghao Wang [email protected] Yang Li [email protected] Zhengyi Jiang [email protected] Guisheng Yin [email protected]
1
College of Computer Science and Technology, Harbin Engineering University, Harbin 150001, China
ation, so it is widely used in many fields, such as seabed survey, ocean mapping, underwater archaeology, and navigation obstacle search. Due to the sensor device and its imaging principle, the sonar image has the problems of low resolution and blurred details when detecting the underwater objects. Therefore, making a high-resolution image from a low-resolution image is an important issue. Because an image with higher resolution usually contains richer details, this can contribute to the execution of subsequent works, such as object reco
Data Loading...
