Abstract:Super-resolution (SR) methods have significantly contributed to the improvement of the spatial resolution of remote sensing (RS) images. The development of deep learning empowers novel methods to learn informative feature representation from massive low-resolution (LR) and high-resolution (HR) image pairs. Conventional RS image SR methods, however, may fail in large-scale (x8 and x9) SR tasks. Specifically, a larger scale factor corresponds to less information in LR images, which is a considerable challenge to SR. To address the issue, we propose a novel method for single RS image SR (SRSISR) based on stratified dense sampling to effectively extract image features. Specifically, the proposed SR dense-sampling residual attention network (SRDSRAN) combines dense sampling and residual learning to improve multilevel feature fusion and gradient propagation and employs local and global attentions to learn important features and long-range interdependence in the channel and spatial dimensions. Meanwhile, we also devise a discriminator model using local and global attentions and with the loss function integrating L(1 )pixel loss, L-1 perceptual loss, and relativistic adversarial loss to obtain the perceptually realistic images. Besides, we introduce a chain training to promote performance and expedite the training process for large-scale SR. Experimental results on UC Merced image and other multispectral data demonstrated that our SRDSRAN outperformed the current state-of-the-art methods quantitatively and in visual quality and obtained a higher classification accuracy in scene classification, proving its potential for applications with other downstream tasks. The code of SRADSGAN will be available at https://github.com/Meng-333/SRADSGAN.

SRAGAN: Generating Colour Landscape Photograph from Sketch.

Attribute-Guided Sketch Generation

Creative and Diverse Artwork Generation Using Adversarial Networks

PencilArt: A Chromatic Penciling Style Generation Framework.

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

Enhancing Urban Landscape Design: A GAN-Based Approach for Rapid Color Rendering of Park Sketches

SRAGAN: Saliency Regularized and Attended Generative Adversarial Network for Chinese Ink-wash Painting Generation

DLP-GAN: learning to draw modern Chinese landscape photos with generative adversarial network

SPDGAN: A Generative Adversarial Network based on SPD Manifold Learning for Automatic Image Colorization

Dual Color Space Guided Sketch Colorization

End-to-End Chinese Landscape Painting Creation Using Generative Adversarial Networks

Sketch to Building: Architecture Image Translation Based on GAN

Picture that Sketch: Photorealistic Image Generation from Abstract Sketches

SCGAN: Saliency Map-guided Colorization with Generative Adversarial Network

Color Sketch Generation

Sgrgan: sketch-guided restoration for traditional Chinese landscape paintings

Towards Criminal Sketching with Generative Adversarial Network

GGD-GAN: Gradient-Guided Dual-Branch Adversarial Networks for Relic Sketch Generation

UGSC‐GAN: User‐guided sketch colorization with deep convolution generative adversarial networks

Face Sketch Synthesis via Semantic-Driven Generative Adversarial Network

SVBRDF Recovery from a Single Image with Highlights Using a Pre-trained Generative Adversarial Network