Self-supervised cycle-consistent learning for scale-arbitrary real-world single image super-resolution

Learning based single image super-resolution (SISR) for real-world images has been an active research topic yet a challenging task, due to the lack of paired low-resolution (LR) and high-resolution (HR) training images. Most of the existing unsupervised real-world SISR methods adopt a two-stage training strategy by synthesizing realistic LR images from their HR counterparts first, then training the super-resolution (SR) models in a supervised manner. However, the training of image degradation and SR models in this strategy are separate, ignoring the inherent mutual dependency between downscaling and its inverse upscaling process. Additionally, the ill-posed nature of image degradation is not fully considered. In this paper, we propose an image downscaling and SR model dubbed as SDFlow, which simultaneously learns a bidirectional many-to-many mapping between real-world LR and HR images unsupervisedly. The main idea of SDFlow is to decouple image content and degradation information in the latent space, where content information distribution of LR and HR images is matched in a common latent space. Degradation information of the LR images and the high-frequency information of the HR images are fitted to an easy-to-sample conditional distribution. Experimental results on real-world image SR datasets indicate that SDFlow can generate diverse realistic LR and SR images both quantitatively and qualitatively.

The single image super resolution recovers missing high resolution details so as to reconstruct a high resolution image from a single low resolution image. This paper proposes a novel directionally adaptive, learning-based, single image super resolution method using multiple direction wavelet transform, called directionlets. Here, critically sampled directionlets are used to capture directional features effectively and to extract edge information along different directions from a set of available high resolution images. This information is used as the training set for super resolving a low resolution input image. The directionlet coefficients at finer scales of its high resolution image are learned locally from this training set and the inverse directionlet transform recovers the super resolved high resolution image. The simulation results showed that the proposed directionlet approach outperforms standard interpolation techniques like cubic spline interpolation as well as standard wavelet-based learning, both visually and in terms of the mean squared error (MSE) values. The SNR scores for cubic spline interpolation, wavelet and directionlet method are 13.6998 dB, 23.8324 dB and 30.8654 dB respectively for Barbara.

Limited resolution is one of the most important factors hindering the application of remote sensing images (RSIs). Single-image super resolution (SISR) is a technique to improve the spatial resolution of digital images and has attracted the attention of many researchers. In recent years, with the advancement of deep learning (DL) frameworks, many DL-based SISR models have been proposed and achieved state-of-the-art performance; however, most SISR models for RSIs use the bicubic downsampler to construct low-resolution (LR) and high-resolution (HR) training pairs. Considering that the quality of the actual RSIs depends on a variety of factors, such as illumination, atmosphere, imaging sensor responses, and signal processing, training on “ideal” datasets results in a dramatic drop in model performance on real RSIs. To address this issue, we propose to build a more realistic training dataset by modeling the degradation with blur kernels and imaging noises. We also design a novel residual balanced attention network (RBAN) as a generator to estimate super-resolution results from the LR inputs. To encourage RBAN to generate more realistic textures, we apply a UNet-shape discriminator for adversarial training. Both referenced evaluations on synthetic data and non-referenced evaluations on actual images were carried out. Experimental results validate the effectiveness of the proposed framework, and our model exhibits state-of-the-art performance in quantitative evaluation and visual quality. We believe that the proposed framework can facilitate super-resolution techniques from research to practical applications in RSIs processing.

The external example-driven single image super-resolution (SISR) method that uses a deep convolutional neural network (CNN) has exhibited superior performance as compared to previously developed SISR methods. However, the advantages of jointly using external and internal examples on a deep CNN framework have not been sufficiently investigated. In this letter, we present a novel method for single image super-resolution by exploiting a complementary relation between external and internal example-based SISR methods. The proposed deep CNN model consists of two subnetworks, a global residual network and a self-residual network, to utilize the advantages of both external and internal examples. In contrast with conventional joint SISR methods, the proposed method is the first deep CNN-based SISR method that does not require a retraining process, which tends to be inefficient. The proposed method outperformed existing methods in both quantitative and qualitative evaluations.

We propose an intuitive formulation for single image super resolution (SISR), and an algorithm based on it that outperforms the state of the art. We model the SISR problem around an aspect of natural images that is often overlooked - sharp edges - which are important for perceptual quality, and troublesome for interpolation. We justify the use of low resolution (LR) Markovian neighborhoods to estimate high resolution (HR) pixels corresponding to only the central pixel of the LR neighborhood based on our formulation. The formulation also lends itself to learning LR to HR mapping based on their pairs as training examples. We propose a learning algorithm based on polynomial neural networks to learn this mapping. Our formulation and algorithm provide further insight into performance of various single image super resolution methods.

Lightweight single image super-resolution with attentive residual refinement network

Sparse representation-based single image super-resolution (SISR) methods use a coupled overcomplete dictionary trained from high-resolution images/image patches. Since remote sensing (RS) satellites capture images of large areas, these images usually have poor spatial resolution and obtaining an effective dictionary as such would be very challenging. Moreover, traditional patch-based sparse representation models for reconstruction tend to give unstable sparse solution and produce visual artefact in the recovered images. To mitigate these problems, in this article, we have proposed an adaptive joint sparse representation-based SISR method that is dependent only on the input low-resolution image for dictionary training and sparse reconstruction. The new model combines patch-based local sparsity and group sparse representation-based nonlocal sparsity in a single framework, which helps in stabilizing the sparse solution and improve the SISR results. The experimental results are evaluated both visually and quantitatively for several RGB and multispectral RS datasets, where the proposed method shows improvements in peak signal-to-noise ratio by 1–4 dB and 2–3 dB over the state-of-the-art sparse representation- and deep learning-based SR methods, respectively. Land cover classification applied on the super-resolved images further validate the advantages of the proposed method. Finally, for practical RS applications, we have performed parallel implementation in general purpose graphics processing units and achieved significant speed ups (30–40×) in the execution time.

Existing single image super-resolution (SISR) methods usually focus on Low-Resolution (LR) images which are artificially generated from High-Resolution (HR) images by a down-sampling process, but are not robust for unmatched training set and testing set. This paper proposes a GAN Flexible Lmser (GFLmser) network that bidirectionally learns the High-to-Low (H2L) process that degrades HR images to LR images and the Low-to-High (L2H) process that recovers the LR images back to HR images. The two directions share the same architecture, added with the gated skip connections from the H2L-net to the L2H-net in order to enhance information transferring for super-resolution. In comparison with several related state-of-the-art methods, experiments demonstrate that not only GFLmser is the most robust method on images of unmatched training set and testing set, but also its performance on real-world face LR images is best in PSNR and reasonably good in FID.

TSRGAN: Real-world text image super-resolution based on adversarial learning and triplet attention

We present a diverse dataset of industrial metal objects with unique characteristics such as symmetry, texturelessness, and high reflectiveness. These features introduce challenging conditions that are not captured in existing datasets. Our dataset comprises both real-world and synthetic multi-view RGB images with 6D object pose labels. Real-world data were obtained by recording multi-view images of scenes with varying object shapes, materials, carriers, compositions, and lighting conditions. This resulted in over 30,000 real-world images. We introduce a new public tool that enables the quick annotation of 6D object pose labels in multi-view images. This tool was used to provide 6D object pose labels for all real-world images. Synthetic data were generated by carefully simulating real-world conditions and varying them in a controlled and realistic way. This resulted in over 500,000 synthetic images. The close correspondence between synthetic and real-world data and controlled variations will facilitate sim-to-real research. Our focus on industrial conditions and objects will facilitate research on computer vision tasks, such as 6D object pose estimation, which are relevant for many industrial applications, such as machine tending. The dataset and accompanying resources are available on the project website.

Deep learning-based single image super-resolution (SISR) methods face various challenges when applied to 3D medical volumetric data (i.e., CT and MR images) due to the high memory cost and anisotropic resolution, which adversely affect their performance. Furthermore, mainstream SISR methods are designed to work over specific upsampling factors, which makes them ineffective in clinical practice. In this paper, we introduce a Spatially Aware Interpolation NeTwork (SAINT) for medical slice synthesis to alleviate the memory constraint that volumetric data poses. Compared to other super-resolution methods, SAINT utilizes voxel spacing information to provide desirable levels of details, and allows for the upsampling factor to be determined on the fly. Our evaluations based on 853 CT scans from four datasets that contain liver, colon, hepatic vessels, and kidneys show that SAINT consistently outperforms other SISR methods in terms of medical slice synthesis quality, while using only a single model to deal with different upsampling factors.

Deep learning can be used to achieve single‐image superresolution (SR) reconstruction. To address problems encountered during this process, such as the number of network parameters, high training requirements on equipment performance, and inability to downsample certain SR images accurately, an image SR reconstruction algorithm based on deep residual network optimization is proposed. The model introduces wavelet transforms based on the original U‐Net, where the U‐Net is trained to obtain SR wavelet feature images at multiple scales simultaneously. This approach reduces the mapping space for the network to learn low‐ to high‐resolution image mapping, which in turn reduces the training difficulty of the model. In terms of network details, the inverse wavelet transform is used in image upsampling to enhance the sparsity of the reconstruction layer in the original network. The network structure of the U‐Net upsampling is adjusted slightly to enable the network to distinguish wavelet images from feature images, thereby improving the richness of the features extracted by the model. The experimental results show that the peak signal‐to‐noise ratio (PSNR) of the fourfold SR model is 32.35 and 28.68 on the Set5 and Set14 validation sets, respectively. Compared with networks that use wavelet prediction mechanisms, such as the deep wavelet prediction SR (DWSR) and deep wavelet prediction‐based residual SR (DWRSR) models, the PSNR for all the tested public datasets is improved by 0.5. The method yields superior results in terms of both visual effect and PSNR, demonstrating the feasibility of the wavelet prediction mechanism in SR reconstruction and thus offering application value and research significance.

Traditional works have shown that patches in a natural image tend to redundantly recur many times inside the image, both within the same scale, as well as across different scales. Make full use of these multi-scale information can improve the image restoration performance. However, the current proposed deep learning based restoration methods do not take the multi-scale information into account. In this paper, we propose a dilated convolution based inception module to learn multi-scale information and design a deep network for single image super-resolution. Different dilated convolution learns different scale feature, then the inception module concatenates all these features to fuse multi-scale information. In order to increase the reception field of our network to catch more contextual information, we cascade multiple inception modules to constitute a deep network to conduct single image super-resolution. With the novel dilated convolution based inception module, the proposed end-to-end single image super-resolution network can take advantage of multi-scale information to improve image super-resolution performance. Experimental results show that our proposed method outperforms many state-of-the-art single image super-resolution methods.

SRDiff: Single image super-resolution with diffusion probabilistic models

Single Image Super-Resolution (SISR) using Convolutional Neural Networks (CNNs) for many applications in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">supervised</i> manner has resulted in state-of-the-art results. Such supervised models achieve remarkable accuracy; albeit their poor generalization ability for real-world Low-Resolution (LR) images. Supervised training in many SR works involves synthetically generated LR images from its corresponding High-Resolution (HR) images. As the distribution of such LR observation is relatively different from that of real LR image, the supervised training in SISR task results degradation when applied on real-world data. SISR has been scaled to real-world data recently by posing the unsupervised problem into a supervised one through learning the distribution of noisy LR observation first, following which supervised training is performed to obtain the SR image. It therefore involves two steps where the accuracy of SR image relies on how closely the LR’s distribution is learnt in the first step. In this work, we overcome such limitation by introducing unsupervised denoising network to transform real noisy LR image to clean image and then pre-trained SR network is utilised to increase the spatial resolution of cleaned LR image to generate SR image. Thus, instead of evaluating the denoised image in LR space to train the denoising network, we inspect the denoised image in SR space which allows to overcome the SR network’s generalization problem. The proposed Unsupervised Denoising framework for Super-Resolution ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">UDSR</i> ) is validated on real-world datasets (NTIRE-2020 Real-World SR Challenge validation and testing dataset (Track-1)) by comparing it with many recent unsupervised SISR methods. The performance of denoising and SR networks is superior in terms of various perceptual indices such as Perceptual Index (PI) and Ma Score in addition to numerous non-references metrics.