3D-CNNHSR: A 3-Dimensional Convolutional Neural Network for Hyperspectral Super-Resolution

Hyperspectral images can easily discriminate different materials due to their fine spectral resolution.However,obtaining a hyperspectral image(HSI)with a high spatial resolution is still a challenge as we are limited by the high computing requirements.The spatial resolution of HSI can be enhanced by utilizing Deep Learning(DL)based Super-resolution(SR).A 3D-CNNHSR model is developed in the present investigation for 3D spatial super-resolution for HSI,without losing the spectral content.The 3DCNNHSR model was tested for the Hyperion HSI.The pre-processing of the HSI was done before applying the SR model so that the full advantage of hyperspectral data can be utilized with minimizing the errors.The key innovation of the present investigation is that it used 3D convolution as it simultaneously applies convolution in both the spatial and spectral dimensions and captures spatial-spectral features.By clustering contiguous spectral content together,a cube is formed and by convolving the cube with the 3D kernel a 3D convolution is realized.The 3D-CNNHSR model was compared with a 2D-CNN model,additionally,the assessment was based on higherresolution data from the Sentinel-2 satellite.Based on the evaluation metrics it was observed that the 3D-CNNHSR model yields better results for the SR of HSI with efficient computational speed,which is significantly less than previous studies.

Shear Let Transform Residual Learning Approach for Single-Image Super-Resolution

Super-resolution techniques are employed to enhance image resolution by reconstructing high-resolution images from one or more low-resolution inputs.Super-resolution is of paramount importance in the context of remote sensing,satellite,aerial,security and surveillance imaging.Super-resolution remote sensing imagery is essential for surveillance and security purposes,enabling authorities to monitor remote or sensitive areas with greater clarity.This study introduces a single-image super-resolution approach for remote sensing images,utilizing deep shearlet residual learning in the shearlet transform domain,and incorporating the Enhanced Deep Super-Resolution network(EDSR).Unlike conventional approaches that estimate residuals between high and low-resolution images,the proposed approach calculates the shearlet coefficients for the desired high-resolution image using the provided low-resolution image instead of estimating a residual image between the high-and low-resolution image.The shearlet transform is chosen for its excellent sparse approximation capabilities.Initially,remote sensing images are transformed into the shearlet domain,which divides the input image into low and high frequencies.The shearlet coefficients are fed into the EDSR network.The high-resolution image is subsequently reconstructed using the inverse shearlet transform.The incorporation of the EDSR network enhances training stability,leading to improved generated images.The experimental results from the Deep Shearlet Residual Learning approach demonstrate its superior performance in remote sensing image recovery,effectively restoring both global topology and local edge detail information,thereby enhancing image quality.Compared to other networks,our proposed approach outperforms the state-of-the-art in terms of image quality,achieving an average peak signal-to-noise ratio of 35 and a structural similarity index measure of approximately 0.9.

Faster split-based feedback network for image super-resolution

Although most of the existing image super-resolution(SR)methods have achieved superior performance,contrastive learning for high-level tasks has not been fully utilized in the existing image SR methods based on deep learning.This work focuses on two well-known strategies developed for lightweight and robust SR,i.e.,contrastive learning and feedback mechanism,and proposes an integrated solution called a split-based feedback network(SPFBN).The proposed SPFBN is based on a feedback mechanism to learn abstract representations and uses contrastive learning to explore high information in the representation space.Specifically,this work first uses hidden states and constraints in recurrent neural network(RNN)to implement a feedback mechanism.Then,use contrastive learning to perform representation learning to obtain high-level information by pushing the final image to the intermediate images and pulling the final SR image to the high-resolution image.Besides,a split-based feedback block(SPFB)is proposed to reduce model redundancy,which tolerates features with similar patterns but requires fewer parameters.Extensive experimental results demonstrate the superiority of the proposed method in comparison with the state-of-the-art methods.Moreover,this work extends the experiment to prove the effectiveness of this method and shows better overall reconstruction quality.

Hyperspectral Image Super-Resolution Meets Deep Learning:A Survey and Perspective

Hyperspectral image super-resolution,which refers to reconstructing the high-resolution hyperspectral image from the input low-resolution observation,aims to improve the spatial resolution of the hyperspectral image,which is beneficial for subsequent applications.The development of deep learning has promoted significant progress in hyperspectral image super-resolution,and the powerful expression capabilities of deep neural networks make the predicted results more reliable.Recently,several latest deep learning technologies have made the hyperspectral image super-resolution method explode.However,a comprehensive review and analysis of the latest deep learning methods from the hyperspectral image super-resolution perspective is absent.To this end,in this survey,we first introduce the concept of hyperspectral image super-resolution and classify the methods from the perspectives with or without auxiliary information.Then,we review the learning-based methods in three categories,including single hyperspectral image super-resolution,panchromatic-based hyperspectral image super-resolution,and multispectral-based hyperspectral image super-resolution.Subsequently,we summarize the commonly used hyperspectral dataset,and the evaluations for some representative methods in three categories are performed qualitatively and quantitatively.Moreover,we briefly introduce several typical applications of hyperspectral image super-resolution,including ground object classification,urban change detection,and ecosystem monitoring.Finally,we provide the conclusion and challenges in existing learning-based methods,looking forward to potential future research directions.

Contrastive Learning for Blind Super-Resolution via A Distortion-Specific Network

Previous deep learning-based super-resolution(SR)methods rely on the assumption that the degradation process is predefined(e.g.,bicubic downsampling).Thus,their performance would suffer from deterioration if the real degradation is not consistent with the assumption.To deal with real-world scenarios,existing blind SR methods are committed to estimating both the degradation and the super-resolved image with an extra loss or iterative scheme.However,degradation estimation that requires more computation would result in limited SR performance due to the accumulated estimation errors.In this paper,we propose a contrastive regularization built upon contrastive learning to exploit both the information of blurry images and clear images as negative and positive samples,respectively.Contrastive regularization ensures that the restored image is pulled closer to the clear image and pushed far away from the blurry image in the representation space.Furthermore,instead of estimating the degradation,we extract global statistical prior information to capture the character of the distortion.Considering the coupling between the degradation and the low-resolution image,we embed the global prior into the distortion-specific SR network to make our method adaptive to the changes of distortions.We term our distortion-specific network with contrastive regularization as CRDNet.The extensive experiments on synthetic and realworld scenes demonstrate that our lightweight CRDNet surpasses state-of-the-art blind super-resolution approaches.

Residual Feature Attentional Fusion Network for Lightweight Chest CT Image Super-Resolution

The diagnosis of COVID-19 requires chest computed tomography(CT).High-resolution CT images can provide more diagnostic information to help doctors better diagnose the disease,so it is of clinical importance to study super-resolution(SR)algorithms applied to CT images to improve the reso-lution of CT images.However,most of the existing SR algorithms are studied based on natural images,which are not suitable for medical images;and most of these algorithms improve the reconstruction quality by increasing the network depth,which is not suitable for machines with limited resources.To alleviate these issues,we propose a residual feature attentional fusion network for lightweight chest CT image super-resolution(RFAFN).Specifically,we design a contextual feature extraction block(CFEB)that can extract CT image features more efficiently and accurately than ordinary residual blocks.In addition,we propose a feature-weighted cascading strategy(FWCS)based on attentional feature fusion blocks(AFFB)to utilize the high-frequency detail information extracted by CFEB as much as possible via selectively fusing adjacent level feature information.Finally,we suggest a global hierarchical feature fusion strategy(GHFFS),which can utilize the hierarchical features more effectively than dense concatenation by progressively aggregating the feature information at various levels.Numerous experiments show that our method performs better than most of the state-of-the-art(SOTA)methods on the COVID-19 chest CT dataset.In detail,the peak signal-to-noise ratio(PSNR)is 0.11 dB and 0.47 dB higher on CTtest1 and CTtest2 at×3 SR compared to the suboptimal method,but the number of parameters and multi-adds are reduced by 22K and 0.43G,respectively.Our method can better recover chest CT image quality with fewer computational resources and effectively assist in COVID-19.

Improved spatiotemporal resolution of anti-scattering super-resolution label-free microscopy via synthetic wave 3D metalens imaging

Super-resolution(SR)microscopy has dramatically enhanced our understanding of biological processes.However,scattering media in thick specimens severely limits the spatial resolution,often rendering the images unclear or indistinguishable.Additionally,live-cell imaging faces challenges in achieving high temporal resolution for fast-moving subcellular structures.Here,we present the principles of a synthetic wave microscopy(SWM)to extract three-dimensional information from thick unlabeled specimens,where photobleaching and phototoxicity are avoided.SWM exploits multiple-wave interferometry to reveal the specimen’s phase information in the area of interest,which is not affected by the scattering media in the optical path.SWM achieves~0.42λ/NA resolution at an imaging speed of up to 106 pixels/s.SWM proves better temporal resolution and sensitivity than the most conventional microscopes currently available while maintaining exceptional SR and anti-scattering capabilities.Penetrating through the scattering media is challenging for conventional imaging techniques.Remarkably,SWM retains its efficacy even in conditions of low signal-to-noise ratios.It facilitates the visualization of dynamic subcellular structures in live cells,encompassing tubular endoplasmic reticulum(ER),lipid droplets,mitochondria,and lysosomes.

SOFFLFM:Super-resolution optical fluctuation Fourierlight-field microscopy

Fourier light-field microscopy(FLFM)uses a microlens aray(MLA)to segment the Fourierplane of the microscopic objective lens to generate multiple two-dimensional perspective views,thereby reconstructing the threedimensional(3D)structure of the sample using 3D deconvo-lution calculation without scanning.However,the resolution of FLFM is stil limited by dif-fraction,and furthermore,it is dependent on the aperture division.In order to improve itsresolution,a super-resolution opticai fuctuation Fourier light-field microscopy(SOFFLFM)wasproposed here,in which the super-resolution optical fluctuation imaging(SOFI)with the abilityof super-resolution was introduced into FLFM.SOFFLFM uses higher-order cumulants statis-tical analysis on an image sequence collected by FLFM,and then carries out 3D deconvolutioncalculation to reconstruct the 3D structure of the sample.The theoretical basis of SOFFLFM onimproving resolution was explained and then verified with the simulations.Simulation resultsdemonstrated that SOFFLFM improved the lateral and axial resolution by more than V2 and 2times in the second-and fourth-order accumulations,compared with that of FLFM.

Super-resolution parameter estimation of monopulse radar by wide-narrowband joint processing

The angular resolution of radar is of crucial signifi-cance to its tracking performance.In this paper,a super-resolu-tion parameter estimation algorithm based on wide-narrowband joint processing is proposed to improve the angular resolution of wideband monopulse radar.The range cells containing resolv-able scattering points are detected in the wideband mode,and these range cells are adopted to estimate part of the target parameters by algorithms of low computational requirement.Then,the likelihood function of the echo is constructed in the narrow-band mode to estimate the rest of the parameters,and the parameters estimated in the wideband mode are employed to reduce computation and enhance estimation accuracy.Simu-lation results demonstrate that the proposed algorithm has higher estimation accuracy and lower computational complexity than the current algorithm and can avoid the risk of model mis-match.

Accelerate Single Image Super-Resolution Using Object Detection Process

Image Super-Resolution(SR)research has achieved great success with powerful neural networks.The deeper networks with more parameters improve the restoration quality but add the computation complexity,which means more inference time would be cost,hindering image SR from practical usage.Noting the spatial distribution of the objects or things in images,a twostage local objects SR system is proposed,which consists of two modules,the object detection module and the SR module.Firstly,You Only Look Once(YOLO),which is efficient in generic object detection tasks,is selected to detect the input images for obtaining objects of interest,then put them into the SR module and output corresponding High-Resolution(HR)subimages.The computational power consumption of image SR is optimized by reducing the resolution of input images.In addition,we establish a dataset,TrafficSign500,for our experiment.Finally,the performance of the proposed system is evaluated under several State-Of-The-Art(SOTA)YOLOv5 and SISR models.Results show that our system can achieve a tremendous computation improvement in image SR.

Using GAN Neural Networks for Super-Resolution Reconstruction of Temperature Fields

A Generative Adversarial Neural(GAN)network is designed based on deep learning for the Super-Resolution(SR)reconstruction task of temperaturefields(comparable to downscaling in the meteorologicalfield),which is limited by the small number of ground stations and the sparse distribution of observations,resulting in a lack offineness of data.To improve the network’s generalization performance,the residual structure,and batch normalization are used.Applying the nearest interpolation method to avoid over-smoothing of the climate element values instead of the conventional Bicubic interpolation in the computer visionfield.Sub-pixel convolution is used instead of transposed convolution or interpolation methods for up-sampling to speed up network inference.The experimental dataset is the European Centre for Medium-Range Weather Forecasts Reanalysis v5(ERA5)with a bidirectional resolution of 0:1°×0:1°.On the other hand,the task aims to scale up the size by a factor of 8,which is rare compared to conventional methods.The comparison methods include traditional interpolation methods and a more widely used GAN-based network such as the SRGAN.Thefinal experimental results show that the proposed scheme advances the performance of Root Mean Square Error(RMSE)by 37.25%,the Peak Signal-to-noise Ratio(PNSR)by 14.4%,and the Structural Similarity(SSIM)by 10.3%compared to the Bicubic Interpolation.For the traditional SRGAN network,a relatively obvious performance improvement is observed by experimental demonstration.Meanwhile,the GAN network can converge stably and reach the approximate Nash equilibrium for various initialization parameters to empirically illustrate the effectiveness of the method in the temperature fields.

A Hybrid Regularization-Based Multi-Frame Super-Resolution Using Bayesian Framework

The prime purpose for the image reconstruction of a multi-frame super-resolution is to reconstruct a higher-resolution image through incorporating the knowledge obtained from a series of relevant low-resolution images,which is useful in numerousfields.Nevertheless,super-resolution image reconstruction methods are usually damaged by undesirable restorative artifacts,which include blurring distortion,noises,and stair-casing effects.Consequently,it is always challenging to achieve balancing between image smoothness and preservation of the edges inside the image.In this research work,we seek to increase the effectiveness of multi-frame super-resolution image reconstruction by increasing the visual information and improving the automated machine perception,which improves human analysis and interpretation processes.Accordingly,we propose a new approach to the image reconstruction of multi-frame super-resolution,so that it is created through the use of the regularization framework.In the proposed approach,the bilateral edge preserving and bilateral total variation regularizations are employed to approximate a high-resolution image generated from a sequence of corresponding images with low-resolution to protect significant features of an image,including sharp image edges and texture details while preventing artifacts.The experimental results of the synthesized image demonstrate that the new proposed approach has improved efficacy both visually and numerically more than other approaches.

Super-Resolution Based on Curvelet Transform and Sparse Representation

Super-resolution techniques are used to reconstruct an image with a high resolution from one or more low-resolution image(s).In this paper,we proposed a single image super-resolution algorithm.It uses the nonlocal mean filter as a prior step to produce a denoised image.The proposed algorithm is based on curvelet transform.It converts the denoised image into low and high frequencies(sub-bands).Then we applied a multi-dimensional interpolation called Lancozos interpolation over both sub-bands.In parallel,we applied sparse representation with over complete dictionary for the denoised image.The proposed algorithm then combines the dictionary learning in the sparse representation and the interpolated sub-bands using inverse curvelet transform to have an image with a higher resolution.The experimental results of the proposed super-resolution algorithm show superior performance and obviously better-recovering images with enhanced edges.The comparison study shows that the proposed super-resolution algorithm outperforms the state-of-the-art.The mean absolute error is 0.021±0.008 and the structural similarity index measure is 0.89±0.08.

IRMIRS:Inception-ResNet-Based Network for MRI Image Super-Resolution

Medical image super-resolution is a fundamental challenge due to absorption and scattering in tissues.These challenges are increasing the interest in the quality of medical images.Recent research has proven that the rapid progress in convolutional neural networks(CNNs)has achieved superior performance in the area of medical image super-resolution.However,the traditional CNN approaches use interpolation techniques as a preprocessing stage to enlarge low-resolution magnetic resonance(MR)images,adding extra noise in the models and more memory consumption.Furthermore,conventional deep CNN approaches used layers in series-wise connection to create the deeper mode,because this later end layer cannot receive complete information and work as a dead layer.In this paper,we propose Inception-ResNet-based Network for MRI Image Super-Resolution known as IRMRIS.In our proposed approach,a bicubic interpolation is replaced with a deconvolution layer to learn the upsampling filters.Furthermore,a residual skip connection with the Inception block is used to reconstruct a high-resolution output image from a low-quality input image.Quantitative and qualitative evaluations of the proposed method are supported through extensive experiments in reconstructing sharper and clean texture details as compared to the state-of-the-art methods.

Deep-learning-based methods for super-resolution fluorescence microscopy

The algorithm used for reconstruction or resolution enhancement is one of the factors affectingthe quality of super-resolution images obtained by fluorescence microscopy.Deep-learning-basedalgorithms have achieved stateof-the-art performance in super-resolution fluorescence micros-copy and are becoming increasingly attractive.We firstly introduce commonly-used deep learningmodels,and then review the latest applications in terms of the net work architectures,the trainingdata and the loss functions.Additionally,we discuss the challenges and limits when using deeplearning to analyze the fluorescence microscopic data,and suggest ways to improve the reliability and robustness of deep learning applications.

RealFuVSR:Feature enhanced real-world video super-resolution

Background Recurrent recovery is a common method for video super-resolution(VSR)that models the correlation between frames via hidden states.However,the application of this structure in real-world scenarios can lead to unsatisfactory artifacts.We found that in real-world VSR training,the use of unknown and complex degradation can better simulate the degradation process in the real world.Methods Based on this,we propose the RealFuVSR model,which simulates real-world degradation and mitigates artifacts caused by the VSR.Specifically,we propose a multiscale feature extraction module(MSF)module that extracts and fuses features from multiple scales,thereby facilitating the elimination of hidden state artifacts.To improve the accuracy of the hidden state alignment information,RealFuVSR uses an advanced optical flow-guided deformable convolution.Moreover,a cascaded residual upsampling module was used to eliminate noise caused by the upsampling process.Results The experiment demonstrates that RealFuVSR model can not only recover high-quality videos but also outperforms the state-of-the-art RealBasicVSR and RealESRGAN models.

Transfer learning-based super-resolution in panoramic models for predicting mandibular third molar extraction difficulty: a multi-center study

Background:This study aims to predict the extraction difficulty of mandibular third molars based on panoramic images using transfer learning while employing super-resolution(SR)technology to enhance the feasibility and validity of the prediction.Methods:We reviewed a total of 608 preoperative mandibular third molar panoramic radiographs from two medical facilities:the First Affiliated Hospital of Zhengzhou University(n=509;456 in the training set and 53 in the test set)and the Henan Provincial Dental Hospital(n=99 in the validation set).We conducted a deep-transfer learning network on high-resolution(HR)panoramic radiographs to improve the longitudinal resolution of the images and obtained the SR images.Subsequently,we constructed models named Model-HR and Model-SR using high-dimensional quantitative features extracted through the Least Absolute Shrinkage and Selection Operator method.The models’performances were evaluated using the receiver operating characteristic curve(ROC).To assess the reliability of the model,we compared the results from the test set with those of three dentists.Results:Model-SR outperformed Model-HR(area under the curve(AUC):0.779,sensitivity:85.5%,specificity:60.9%,and accuracy:79.8%vs.AUC:0.753,sensitivity:73.7%,specificity:73.9%,and accuracy:73.7%)in predicting the difficulty of extracting mandibular third molars.Both Model-HR(AUC=0.821,95%CI 0.687–0.956)and Model-SR(AUC=0.963,95%CI 0.921–0.999)demonstrated superior performance compared to expert dentists(highest AUC=0.799,95%CI 0.671–0.927).Conclusions:Model-SR yielded superior predictive performance in determining the difficulty of extracting mandibular third molars when compared with Model-HR and expert dentists’visual assessments.

Decoupling of temporal/spatial broadening effects in Doppler wind LiDAR by 2D spectral analysis

Pulse echo accumulation is commonly employed in coherent Doppler wind LiDAR(light detection and ranging)under the assumption of steady wind.Here,the measured spectral data are analyzed in the time dimension and frequency dimension to cope with the temporal wind shear and achieve the optimal accumulation time.A hardware-efficient algorithm combining the interpolation and cross-correlation is used to enhance the wind retrieval accuracy by reducing the frequency sampling interval and then reduce the spectral width calculation error.Moreover,the temporal broadening effect and spatial broadening effect are decoupled according to the strategy we developed.

Preoperative prediction of lymphovascular and perineural invasion in gastric cancer using spectral computed tomography imaging and machine learning

BACKGROUND Lymphovascular invasion(LVI)and perineural invasion(PNI)are important prognostic factors for gastric cancer(GC)that indicate an increased risk of metastasis and poor outcomes.Accurate preoperative prediction of LVI/PNI status could help clinicians identify high-risk patients and guide treatment deci-sions.However,prior models using conventional computed tomography(CT)images to predict LVI or PNI separately have had limited accuracy.Spectral CT provides quantitative enhancement parameters that may better capture tumor invasion.We hypothesized that a predictive model combining clinical and spectral CT parameters would accurately preoperatively predict LVI/PNI status in GC patients.AIM To develop and test a machine learning model that fuses spectral CT parameters and clinical indicators to predict LVI/PNI status accurately.METHODS This study used a retrospective dataset involving 257 GC patients(training cohort,n=172;validation cohort,n=85).First,several clinical indicators,including serum tumor markers,CT-TN stages and CT-detected extramural vein invasion(CT-EMVI),were extracted,as were quantitative spectral CT parameters from the delineated tumor regions.Next,a two-step feature selection approach using correlation-based methods and information gain ranking inside a 10-fold cross-validation loop was utilized to select informative clinical and spectral CT parameters.A logistic regression(LR)-based nomogram model was subsequently constructed to predict LVI/PNI status,and its performance was evaluated using the area under the receiver operating characteristic curve(AUC).RESULTS In both the training and validation cohorts,CT T3-4 stage,CT-N positive status,and CT-EMVI positive status are more prevalent in the LVI/PNI-positive group and these differences are statistically significant(P<0.05).LR analysis of the training group showed preoperative CT-T stage,CT-EMVI,single-energy CT values of 70 keV of venous phase(VP-70 keV),and the ratio of standardized iodine concentration of equilibrium phase(EP-NIC)were independent influencing factors.The AUCs of VP-70 keV and EP-NIC were 0.888 and 0.824,respectively,which were slightly greater than those of CT-T and CT-EMVI(AUC=0.793,0.762).The nomogram combining CT-T stage,CT-EMVI,VP-70 keV and EP-NIC yielded AUCs of 0.918(0.866-0.954)and 0.874(0.784-0.936)in the training and validation cohorts,which are significantly higher than using each of single independent factors(P<0.05).CONCLUSION The study found that using portal venous and EP spectral CT parameters allows effective preoperative detection of LVI/PNI in GC,with accuracy boosted by integrating clinical markers.

Estimating wheat spike-leaf composite indicator(SLI)dynamics by coupling spectral indices and machine learning

The contribution of spike photosynthesis to grain yield(GY)has been overlooked in the accurate spectral prediction of yield.Thus,it’s essential to construct and estimate a yield-related phenotypic trait considering spike photosynthesis.Based on field and spectral reflectance data from 19 wheat cultivars under two nitrogen fertilization conditions in two years,our objectives were to(i)construct a yield-related phenotypic trait(spike–leaf composite indicator,SLI)accounting for the contribution of the spike to photosynthesis,(ii)develop a novel spectral index(enhanced triangle vegetation index,ETVI3)sensitive to SLI,and(iii)establish and evaluate SLI estimation models by integrating spectral indices and machine learning algorithms.The results showed that SLI was sensitive to nitrogen fertilizer and wheat cultivar variation as well as a better predictor of yield than the leaf area index.ETVI3 maintained a strong correlation with SLI throughout the growth stage,whereas the correlations of other spectral indices with SLI were poor after spike emergence.Integrating spectral indices and machine learning algorithms improved the estimation accuracy of SLI,with the most accurate estimates of SLI showing coefficient of determination,root mean square error(RMSE),and relative RMSE values of 0.71,0.047,and 26.93%,respectively.These results provide new insights into the role of fruiting organs for the accurate spectral prediction of GY.This high-throughput SLI estimation approach can be applied for wheat yield prediction at whole growth stages and may be assisted with agronomical practices and variety selection.