Image Super-Resolution Reconstruction Based on a Generative Adversarial Network

Abstract

In the field of computer vision, super-resolution reconstruction techniques based on deep learning have undergone considerable advancement; however, certain limitations remain, such as insufficient feature extraction and blurred image generation. To address these problems, we propose an image super-resolution reconstruction model based on a generative adversarial network. First, we employ a dual network structure in the generator network to solve the problem of insufficient feature extraction. The dual network structure is divided into an upsample subnetwork and a refinement subnetwork, which upsample and optimize a low-resolution image, respectively. In a scene with large upscaling factors, this structure can reduce the negative effect of noise and enhance the utilization of high-frequency details, thereby generating high-quality reconstruction results. Second, to generate sharper super-resolution images, we use the perceptual loss, which exhibits a fast convergence and excellent visual effect, to guide the generator network training. We apply the ResNeXt-50-32x4d network, which has few parameters and a large depth, to calculate the loss to obtain a reconstructed super-resolution image that is highly realistic. Finally, we introduce the Wasserstein distance into the discriminator network to enhance the discrimination ability and stability of the model. Specifically, this distance is employed to eliminate the activation function in the last layer of the network and avoid the use of the logarithm in calculating the loss function. Extensive experiments on the DIV2K, Set5, Set14, and BSD100 datasets demonstrate the effectiveness of the proposed model.

An Extensive Comparative Analysis on Various Efficient Techniques for Image Super-Resolution

“…Y. Wu et al, [12] propose a picture super-resolution recreation model in light of a generative ill-disposed network. To begin with, we utilize a double organization structure in the generator organization to tackle the issue of deficient component extraction.…”

Mars Image Super-Resolution Based on Generative Adversarial Network

“…Generative Adversarial Network. Recently, GAN [32] have attracted widespread attention [33]- [36]. The essence of GAN is to generate similar distributions through adversarial learning strategy.…”

Image unmosaicing without location information using stacked GAN

“…GAN utilises adversarial training to learn the generator and discriminator alternatively and has shown powerful ability to generate natural images [16, 17]. Owing to GAN's ability to generate images, it has widely been used for research problems such as super resolution [3, 4], texture synthesis [18, 19], domain translation [20, 21], and image completion [9–14, 22, 23].…”

“…estimating a high‐resolution image from its low‐resolution counterpart by exploiting deep residual network with skip connection. Super‐resolution perceptual GAN [4] exploited a classification network and features obtained by the discriminator network to generate a robust perceptual loss. Although Ledig et al [3], Wu et al [4], and other methods work well for image super resolution, they suffer in image unmosaicing due to the reasons described earlier.…”

“…The most attractive feature of our method is that unlike previous image completion methods, it recovers mosaiced parts without their explicit location information. One might think that super‐resolution methods [3, 4] can solve this problem. However, they do not work for image unmosaicing.…”