Low-resolution Cases Research Articles

Enhanced satellite image resolution with a residual network and correlation filter

This study addresses the predominant challenge of very low-resolution satellite images in remote sensing applications, a common issue in satellite image-based surveillance. Existing satellite image recognition algorithms struggle with such low-resolution images, and traditional Super-Resolution (SR) techniques fall short for very low-resolution cases. We propose the Progressive Satellite Image Super-Resolution (PSISR) model to bridge this gap. Unlike current learning-based SR methods, the PSISR model specifically targets very low-resolution satellite images. In satellite image super-resolution, problems with feature fusion that result in image noise, blind spots, poor perceptual quality, and checkboard artifacts are encountered during the reconstruction process. Current models try to improve perceptual quality, but they frequently show challenges in attaining acceptable outcomes because of losses during reconstruction. Using a combined loss function, correlation filters, and a loss-aware upscaling network layer, the PSISR model presents a revolutionary methodology. The model adopts a cascading structure with dense skip connections, sequentially upscaling images by factors of 2×, 4×, and 8× through three modules. To validate the model's superiority, a study is conducted, confirming its effectiveness compared to baseline models and also training the other models using the available dataset to prove the effectiveness of the model. The PSISR model effectively addresses the challenge of extracting more features with minimal losses, resulting in high magnification during reconstruction. Our method outperforms state-of-the-art techniques, including Swin2-MoSE, MambaFormer, SRFBN and RCAN, with a PSNR improvement of up to 0.4 dB and a 0.003 SSIM enhancement across various datasets. This demonstrates the effectiveness of our approach in producing high-quality outputs, achieving a 99.25 % correlation efficiency between the generated and original images.

Resolution Enhancement of SOHO/MDI Magnetograms

Research on the solar magnetic field and its effects on solar dynamo mechanisms and space weather events has benefited from the continual improvements in instrument resolution and measurement frequency. The augmentation and assimilation of historical observational data timelines also play a significant role in understanding the patterns of solar magnetic field variation. Within the realm of astronomical data processing, super-resolution (SR) reconstruction refers to the process of using a substantial corpus of training data to learn the nonlinear mapping between low-resolution (LR) and high-resolution (HR) images, thereby achieving higher-resolution astronomical images. This paper is an application study in high-dimensional nonlinear regression. Deep learning models were employed to perform SR modeling on SOHO/MDI magnetograms and SDO/HMI magnetograms, thus reliably achieving resolution enhancement of full-disk SOHO/MDI magnetograms and enhancing the image resolution to obtain more detailed information. For this study, a data set comprising 9717 pairs of data from 2010 April to 2011 February was used as the training set, 1332 pairs from 2011 March were used as the validation set and 1034 pairs from 2011 April were used as the test set. After data preprocessing, we randomly cropped 128 × 128 sub-images as the LR cases from the full-disk MDI magnetograms, and the corresponding 512 × 512 sub-images as HR ones from the HMI full-disk magnetograms for model training. The tests conducted have shown that the study successfully produced reliable 4× SR reconstruction of full-disk MDI magnetograms. The MESR model’s results (0.911) were highly correlated with the target HMI magnetographs as indicated by the correlation coefficient values. Furthermore, the method achieved the best PSNR, SSIM, MAE and RMSE values, indicating that the MESR model can effectively reconstruct magnetograms.

Multi-Scale Feature Fusion and Structure-Preserving Network for Face Super-Resolution

Deep convolutional neural networks have demonstrated significant performance improvements in face super-resolution tasks. However, many deep learning-based approaches tend to overlook the inherent structural information and feature correlation across different scales in face images, making the accurate recovery of face structure in low-resolution cases challenging. To address this, this paper proposes a method that fuses multi-scale features while preserving the facial structure. It introduces a novel multi-scale residual block (MSRB) to reconstruct key facial parts and structures from spatial and channel dimensions, and utilizes pyramid attention (PA) to exploit non-local self-similarity, improving the details of the reconstructed face. Feature Enhancement Modules (FEM) are employed in the upscale stage to refine and enhance current features using multi-scale features from previous stages. The experimental results on CelebA, Helen and LFW datasets provide evidence that our method achieves superior quantitative metrics compared to the baseline, the Peak Signal-to-Noise Ratio (PSNR) outperforms the baseline by 0.282 dB, 0.343 dB, and 0.336 dB. Furthermore, our method demonstrates improved visual performance on two additional no-reference datasets, Widerface and Webface.

Dual Polarization Modality Fusion Network for Assisting Pathological Diagnosis.

Polarization imaging is sensitive to sub-wavelength microstructures of various cancer tissues, providing abundant optical characteristics and microstructure information of complex pathological specimens. However, how to reasonably utilize polarization information to strengthen pathological diagnosis ability remains a challenging issue. In order to take full advantage of pathological image information and polarization features of samples, we propose a dual polarization modality fusion network (DPMFNet), which consists of a multi-stream CNN structure and a switched attention fusion module for complementarily aggregating the features from different modality images. Our proposed switched attention mechanism could obtain the joint feature embeddings by switching the attention map of different modality images to improve their semantic relatedness. By including a dual-polarization contrastive training scheme, our method can synthesize and align the interaction and representation of two polarization features. Experimental evaluations on three cancer datasets show the superiority of our method in assisting pathological diagnosis, especially in small datasets and low imaging resolution cases. Grad-CAM visualizes the important regions of the pathological images and the polarization images, indicating that the two modalities play different roles and allow us to give insightful corresponding explanations and analysis on cancer diagnosis conducted by the DPMFNet. This technique has potential to facilitate the performance of pathological aided diagnosis and broaden the current digital pathology boundary based on pathological image features.

Comparative analysis of super-resolution reconstructed images for micro-expression recognition

It is an established fact that the genuineness of facial micro-expression is an effective means for estimating concealed emotions (Li et al. in Micro-expression recognition under low-resolution cases. SciTePress, Science and Technology Publications, Setúbal, 2019). Conventionally, analysis of these expressions has been performed using high resolution images which are ideal cases. However, in a real-world scenario, capturing expressions with high resolution images may not always be possible particularly using low-cost surveillance cameras. Faces captured using such cameras are often very tiny and of poor resolution. Due to the loss of discriminative features these images may not be of much use particularly for identifying certain minute facial details. To make these images useful, enhancing the textural information becomes essential and super-resolution algorithms can be ideal to achieve this. In this work, we utilize algorithms based on deep learning and generative adversarial network for transforming low-resolution micro-expression images into super-resolution images and examine their fitness particularly for micro-expression recognition. The proposed approach is tested on simulated dataset obtained from two popular spontaneous micro-expression datasets namely CASME II and SMIC-VIS; the experimental results demonstrate that the method achieved favourable results with the best recognition performance recorded as 61.63%. The significance of this work is: first, it thoroughly investigates reconstruction performance of several deep learning super-resolution algorithms on simulated low-quality micro-expression images; second, it provides a comprehensive analysis of the results obtained employing these reconstructed images to determine their contribution in addressing image quality issues specifically for micro-expression recognition.

Optimization of Different Deployments and Resolution of Antenna Arrays in SWIPT

In this paper, three different deployment antenna arrays with circular, triangular and rectangular shapes were used to optimize the simultaneous wireless information and power transfer (SWIPT) system for the Internet of Things (IoT). Ray-tracing was employed to channel the model for a real environment. Self-adaptive dynamic differential evolution (SADDE) was used to optimize the harvesting power ratio with bit error rate constrained by the two different resolutions of feed length (high resolution and low resolution). Numerical results show that those three antenna arrays can achieve the goal for information quality in both resolutions. The harvesting power ratio for the circular array is the best and the harvesting power ratio for the rectangular array is the worst. The harvesting power ratio for the low-resolution case is 25% lower than the high-resolution case. However, the circular antenna array is the best deployment in those three different arrays for both high and low resolutions.

Prediction of heat transfer from a circular cylinder in a cross-flow at a low sub-critical Reynolds number with the Partially-Averaged Navier-Stokes method

The prediction of turbulent heat transfer at a reasonable computational cost is a topical issue in computational fluid dynamics. This study addresses the numerical simulation of heat transfer from a circular cylinder in a cross-flow. Simulations were performed at a low sub-critical Reynolds number (Re = 3900) using the Partially-Averaged Navier-Stokes (PANS). Seven ratios of unresolved to total turbulent kinetic energy fk ranging from 0.15 to 0.7 were considered in order to reveal the influence of physical resolution on the flow and heat transfer statistics. Further, a low Reynolds number approach fε = fk was evaluated for fk ≥ 0.5. Additionally, simulations using the Large Eddy Simulation (LES) method were conducted on the same computational grids. The low physical resolution (fk > 0.5) in the PANS simulation at the coarse spatial resolution led to the earlier transition and later separation of the boundary layer which resulted in the overprediction of heat transfer at the rear stagnation point. The low Re approach (fε = fk) provided similar results as the low physical resolution case (fk = 0.7, fε = 1) and spatial refinement did not influence the predicted results. The results of our study indicated the suitability of the PANS for applications involving heat transfer from a bluff body provided an adequate physical resolution level is set.

Super-Resolution Enhanced Medical Image Diagnosis With Sample Affinity Interaction.

The degradation in image resolution harms the performance of medical image diagnosis. By inferring high-frequency details from low-resolution (LR) images, super-resolution (SR) techniques can introduce additional knowledge and assist high-level tasks. In this paper, we propose a SR enhanced diagnosis framework, consisting of an efficient SR network and a diagnosis network. Specifically, a Multi-scale Refined Context Network (MRC-Net) with Refined Context Fusion (RCF) is devised to leverage global and local features for SR tasks. Instead of learning from scratch, we first develop a recursive MRC-Net with temporal context, and then propose a recursion distillation scheme to enhance the performance of MRC-Net from the knowledge of the recursive one and reduce the computational cost. The diagnosis network jointly utilizes the reliable original images and more informative SR images by two branches, with the proposed Sample Affinity Interaction (SAI) blocks at different stages to effectively extract and integrate discriminative features towards diagnosis. Moreover, two novel constraints, sample affinity consistency and sample affinity regularization, are devised to refine the features and achieve the mutual promotion of these two branches. Extensive experiments of synthetic and real LR cases are conducted on wireless capsule endoscopy and histopathology images, verifying that our proposed method is significantly effective for medical image diagnosis.

Super resolution convolutional neural network based pre-processing for automatic polyp detection in colonoscopy images

Colonoscopy is the most common methodology used to detect polyps on the colon surface. Increasing the image resolution has the potential to improve the automatic colonoscopy based diagnosis and polyp detection and localization. In this study, we proposed a pre-processing approach that uses convolutional neural network based super resolution method (SRCNN) to increase the resolution of the training colonoscopy images before the localization of polyps. We also investigated the use of CNN based models such as the Single Shot MultiBox Detector (SSD) and Faster Regional CNN (RCNN) for real-time polyp detection and localization. Our results showed that using SRCNN method before the training process provides better results in terms of accuracy in both models compared to the low-resolution cases. Furthermore, we reached an F2 score of 0.945 for the correct localization of colon polyps using Faster RCNN with ResNet-101 feature extractor.

Image fusion in remote sensing by multi-objective deep learning

ABSTRACT This paper presents a multi-objective deep learning method for the purpose of achieving image fusion or pansharpening in remote sensing. This method uses a Denoising Autoencoder (DA). Two terms are added to the commonly used mean squared error loss function. The first term involves first applying a fractional-order superimposed gradient to the PANchromatic (PAN) image for extracting the high frequency information, and then by considering the difference between the network output and the edge map of the PANchromatic image. The second term involves the difference between the universal image quality index of the low resolution MultiSpectral (MS) image and the network output for each spectral band. These terms allow both the spatial and spectral information to be better preserved in the fused image. Experimental results on three public domain datasets for both low resolution and full resolution cases are reported based on commonly used objective metrics. Compared to the existing methods, the results obtained indicate the developed multi-objective method generates comparable results for the low resolution scenario and superior performance for the full resolution scenario.

Reducing Quantizer Distortion Due to Insufficient Resolution in Massive MIMO Receivers

Use of low-resolution (1-4 bits) Analog-to-Digital Converters (ADCs) can reduce power consumption in Massive Multiple-Input, Multiple-Output (MIMO) receivers. Ordinary linear beamforming may suffice for low-resolution ADCs under conditions on the Signal-to-Noise Ratio (SNR) and number of antennas that may be called low but sufficient resolution. However, if the SNR increases or number of antennas decreases, an error floor will typically occur. We introduce three low-complexity iterative algorithms to reduce quantization noise in such low but insufficient resolution cases. These algorithms process the raw quantizer outputs prior to detection, achieving up to two orders-of-magnitude reduction in Bit Error Rate (BER). Our algorithms are based on the new `equivalent model' for quantizers developed in our prior work. These algorithms can be applied to any number of bits and any modulation format. We focus on Orthogonal Frequency-Division Multiplexing (OFDM) to show that quantizer distortion can be corrected without going to the frequency domain.

Integrating soft and hard dose-volume constraints into hierarchical constrained IMRT optimization.

Dose-volume constraints (DVCs) continue to be common features in intensity-modulated radiation therapy (IMRT) prescriptions, but they are non-convex and difficult to incorporate. We propose computationally efficient methods to incorporate dose-volume constraints (DVCs) into automated IMRT planning. We propose a two-phase approach: in phase-1, we solve a convex approximation with DVCs. Although this convex approximation does not guarantee DVC satisfaction, it provides crucial initial information about voxels likely to receive doses below DVC thresholds. Subsequently, phase-2 solves an optimization problem with maximum dose constraints imposed on those subthreshold voxels. We further categorize DVCs into hard- and soft-DVCs, where hard-DVCs are strictly enforced by the optimization and soft-DVCs are encouraged in the objective function. We tested this approach in our automated treatment planning system which is based on hierarchical constrained optimization. Performance is demonstrated on a series of paraspinal, lung, oligometastasis, and prostate cases as well as a small paraspinal case for which we can computationally afford to obtain a ground-truth by solving a non-convex optimization problem. The proposed algorithm successfully meets all the hard-DVCs while increasing the overall computational time of the baseline planning process (without DVCs) by 20%, 10%, and 11% for paraspinal, oligometastasis, and prostate cases, respectively. For a soft-DVC applied to the lung case, the dose-volume histogram curve moves toward the desired direction and the computational time is increased by 11%. For a low-resolution paraspinal case, the ground-truth solution process using mixed-integer programming methods required 15h while the proposed algorithm converges in only 2min with a proximal solution. A computationally tractable algorithm to handle hard- and soft-DVCs is developed which is capable of satisfying DVCs without any parameter tweaking. Although the algorithm is demonstrated in our in-house developed automated treatment planning system, it can potentially be used in any constrained optimization framework.

Data assimilation using adaptive, non-conservative, moving mesh models

Abstract. Numerical models solved on adaptive moving meshes have become increasingly prevalent in recent years. Motivating problems include the study of fluids in a Lagrangian frame and the presence of highly localized structures such as shock waves or interfaces. In the former case, Lagrangian solvers move the nodes of the mesh with the dynamical flow; in the latter, mesh resolution is increased in the proximity of the localized structure. Mesh adaptation can include remeshing, a procedure that adds or removes mesh nodes according to specific rules reflecting constraints in the numerical solver. In this case, the number of mesh nodes will change during the integration and, as a result, the dimension of the model's state vector will not be conserved. This work presents a novel approach to the formulation of ensemble data assimilation (DA) for models with this underlying computational structure. The challenge lies in the fact that remeshing entails a different state space dimension across members of the ensemble, thus impeding the usual computation of consistent ensemble-based statistics. Our methodology adds one forward and one backward mapping step before and after the ensemble Kalman filter (EnKF) analysis, respectively. This mapping takes all the ensemble members onto a fixed, uniform reference mesh where the EnKF analysis can be performed. We consider a high-resolution (HR) and a low-resolution (LR) fixed uniform reference mesh, whose resolutions are determined by the remeshing tolerances. This way the reference meshes embed the model numerical constraints and are also upper and lower uniform meshes bounding the resolutions of the individual ensemble meshes. Numerical experiments are carried out using 1-D prototypical models: Burgers and Kuramoto–Sivashinsky equations and both Eulerian and Lagrangian synthetic observations. While the HR strategy generally outperforms that of LR, their skill difference can be reduced substantially by an optimal tuning of the data assimilation parameters. The LR case is appealing in high dimensions because of its lower computational burden. Lagrangian observations are shown to be very effective in that fewer of them are able to keep the analysis error at a level comparable to the more numerous observers for the Eulerian case. This study is motivated by the development of suitable EnKF strategies for 2-D models of the sea ice that are numerically solved on a Lagrangian mesh with remeshing.

Approaching Earth’s core conditions in high-resolution geodynamo simulations

SUMMARY The geodynamo features a broad separation between the large scale at which Earth’s magnetic field is sustained against ohmic dissipation and the small scales of the turbulent and electrically conducting underlying fluid flow in the outer core. Here, the properties of this scale separation are analysed using high-resolution numerical simulations that approach closer to Earth’s core conditions than earlier models. The new simulations are obtained by increasing the resolution and gradually relaxing the hyperdiffusive approximation of previously published low-resolution cases. This upsizing process does not perturb the previously obtained large-scale, leading-order quasi-geostrophic (QG) and first-order magneto-Archimedes-Coriolis (MAC) force balances. As a result, upsizing causes only weak transients typically lasting a fraction of a convective overturn time, thereby demonstrating the efficiency of this approach to reach extreme conditions at reduced computational cost. As Earth’s core conditions are approached in the upsized simulations, Ohmic losses dissipate up to 97 per cent of the injected convective power. Kinetic energy spectra feature a gradually broadening self-similar, power-law spectral range extending over more than a decade in length scale. In this range, the spectral energy density profile of vorticity is shown to be approximately flat between the large scale at which the magnetic field draws its energy from convection through the QG-MAC force balance and the small scale at which this energy is dissipated. The resulting velocity and density anomaly planforms in the physical space consist in large-scale columnar sheets and plumes, respectively, co-existing with small-scale vorticity filaments and density anomaly ramifications. In contrast, magnetic field planforms keep their large-scale structure after upsizing. The small-scale vorticity filaments are aligned with the large-scale magnetic field lines, thereby minimizing the dynamical influence of the Lorentz force. The diagnostic outputs of the upsized simulations are more consistent with the asymptotic QG-MAC theory than those of the low-resolution cases that they originate from, but still feature small residual deviations that may call for further theoretical refinements to account for the structuring constraints of the magnetic field on the flow.

An Ecological Land Cover Sampling Reclassification Model for Safety Estimation of Shoreline Systems from a Flood Defense Perspective Using Optical Satellite Remote Sensing Imaging

The safety level of a shoreline is essential for flood control projects and policy formulation or modification from both economic and environmental perspectives. With the development of remote sensing (RS) techniques, high spatial-spectral resolution and quick-revolution satellite images are now available and widely used in environment monitoring and management. It is therefore possible to more efficiently and conveniently identify the components of, and extract information for, shoreline environments. However, the problem is that the shoreline is always a long curve with a relatively narrow width, which limits the application of RS technology. This paper presents a method of recognizing different types of shoreline and of conveniently extracting the geographical coordinates of potential shoreline defense by analyzing and processing ecological information from an optical satellite RS data interpretation of land cover on both side of the shoreline. An application of this model in a low-resolution image case proved that the model can be used in the primary survey of a shoreline monitoring service platform as the basic tile level. The classification model is designed such that the requirements of image resolution for efficiently extracting information from the shoreline are low and the limitations imposed by a narrow shoreline width are avoided.

Hybrid parallelization of the XTOR-2F code for the simulation of two-fluid MHD instabilities in tokamaks

A hybrid MPI/OpenMP parallel version of the XTOR-2F code [Lütjens and Luciani, J. Comput. Phys. 229 (2010) 8130] solving the two-fluid MHD equations in full tokamak geometry by means of an iterative Newton–Krylov matrix-free method has been developed. The present work shows that the code has been parallelized significantly despite the numerical profile of the problem solved by XTOR-2F, i.e. a discretization with pseudo-spectral representations in all angular directions, the stiffness of the two-fluid stability problem in tokamaks, and the use of a direct LU decomposition to invert the physical pre-conditioner at every Krylov iteration of the solver. The execution time of the parallelized version is an order of magnitude smaller than the sequential one for low resolution cases, with an increasing speedup when the discretization mesh is refined. Moreover, it allows to perform simulations with higher resolutions, previously forbidden because of memory limitations.

Automated refinement of macromolecular structures at low resolution using prior information.

Since the ratio of the number of observations to adjustable parameters is small at low resolution, it is necessary to use complementary information for the analysis of such data. ProSMART is a program that can generate restraints for macromolecules using homologous structures, as well as generic restraints for the stabilization of secondary structures. These restraints are used by REFMAC5 to stabilize the refinement of an atomic model. However, the optimal refinement protocol varies from case to case, and it is not always obvious how to select appropriate homologous structure(s), or other sources of prior information, for restraint generation. After running extensive tests on a large data set of low-resolution models, the best-performing refinement protocols and strategies for the selection of homologous structures have been identified. These strategies and protocols have been implemented in the Low-Resolution Structure Refinement (LORESTR) pipeline. The pipeline performs auto-detection of twinning and selects the optimal scaling method and solvent parameters. LORESTR can either use user-supplied homologous structures, or run an automated BLAST search and download homologues from the PDB. The pipeline executes multiple model-refinement instances using different parameters in order to find the best protocol. Tests show that the automated pipeline improves R factors, geometry and Ramachandran statistics for 94% of the low-resolution cases from the PDB included in the test set.

Weak Environmental Controls of Tropical Forest Canopy Height in the Guiana Shield

Canopy height is a key variable in tropical forest functioning and for regional carbon inventories. We investigate the spatial structure of the canopy height of a tropical forest, its relationship with environmental physical covariates, and the implication for tropical forest height variation mapping. Making use of high-resolution maps of LiDAR-derived Digital Canopy Model (DCM) and environmental covariates from a Digital Elevation Model (DEM) acquired over 30,000 ha of tropical forest in French Guiana, we first show that forest canopy height is spatially correlated up to 2500 m. Forest canopy height is significantly associated with environmental variables, but the degree of correlation varies strongly with pixel resolution. On the whole, bottomland forests generally have lower canopy heights than hillslope or hilltop forests. However, this global picture is very noisy at local scale likely because of the endogenous gap-phase forest dynamic processes. Forest canopy height has been predictively mapped across a pixel resolution going from 6 m to 384 m mimicking a low resolution case of 3 points·km − 2 . Results of canopy height mapping indicated that the error for spatial model with environment effects decrease from 8.7 m to 0.91 m, depending of the pixel resolution. Results suggest that, outside the calibration plots, the contribution of environment in shaping the global canopy height distribution is quite limited. This prevents accurate canopy height mapping based only on environmental information, and suggests that precise canopy height maps, for local management purposes, can only be obtained with direct LiDAR monitoring.

Progress in low-resolution ab initio phasing with CrowdPhase.

Ab initio phasing by direct computational methods in low-resolution X-ray crystallography is a long-standing challenge. A common approach is to consider it as two subproblems: sampling of phase space and identification of the correct solution. While the former is amenable to a myriad of search algorithms, devising a reliable target function for the latter problem remains an open question. Here, recent developments in CrowdPhase, a collaborative online game powered by a genetic algorithm that evolves an initial population of individuals with random genetic make-up (i.e. random phases) each expressing a phenotype in the form of an electron-density map, are presented. Success relies on the ability of human players to visually evaluate the quality of these maps and, following a Darwinian survival-of-the-fittest concept, direct the search towards optimal solutions. While an initial study demonstrated the feasibility of the approach, some important crystallographic issues were overlooked for the sake of simplicity. To address these, the new CrowdPhase includes consideration of space-group symmetry, a method for handling missing amplitudes, the use of a map correlation coefficient as a quality metric and a solvent-flattening step. Performances of this installment are discussed for two low-resolution test cases based on bona fide diffraction data.

Optimal quantisation for random parameter estimation

In this study, the optimal quantiser design for random parameter estimation is investigated. The objective is to find a quantiser to minimise the variance of the estimation error by the minimum mean-square estimation. The main results are presented for the cases of high and low resolutions, respectively. For high resolution, multi-dimensional quantisation is considered and a quantitative relationship between the quantisation density and the probability density function is presented. For low-resolution case, an indirect method is developed for one-dimensional optimal quantisation by exploiting the results of high resolution case. The measurement space is first evenly divided into a number of small intervals, then the quantisation is approximately represented by the grouping of the small intervals. At last, a dynamic programming-based method is presented for the optimal grouping.