Journal: Scientific Data

Article Title: A Real-World Benchmark for Sentinel-2 Multi-Image Super-Resolution

doi: 10.1038/s41597-023-02538-9

Figure Lengend Snippet: Reconstruction accuracy obtained for our MuS2 benchmark measured with PSNR (in dB), SSIM, LPIPS, and the balanced score \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr{B}}$$\end{document} B , obtained for different interpolation techniques alongside HighRes-net and RAMS networks trained using real-world and simulated images.

Article Snippet: However, it can be noticed that these qualitative observations are quantitatively reflected only in the LPIPS values, while PSNR and SSIM are slightly worse for HighRes-net and RAMS–overall, both LPIPS and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr{B}}$$\end{document} B indicate that SR networks perform better than interpolation and they penalize for the artefacts.

Techniques:

Journal: Scientific Data

Article Title: A Real-World Benchmark for Sentinel-2 Multi-Image Super-Resolution

doi: 10.1038/s41597-023-02538-9

Figure Lengend Snippet: Reconstruction outcome (band B08) obtained with RAMS and HighRes-net trained from real-life PROBA-V NIR and Red images and from simulated data, compared with image interpolation techniques and HR reference. PSNR (in dB), SSIM, LPIPS and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr{B}}$$\end{document} B scores are presented above each example. The images present an area of 5.5 × 6.5 km near Wageningen, Netherlands–the WV-2 image was acquired in June 2011, and S-2 images were acquired in April 2019–March 2021.

Article Snippet: However, it can be noticed that these qualitative observations are quantitatively reflected only in the LPIPS values, while PSNR and SSIM are slightly worse for HighRes-net and RAMS–overall, both LPIPS and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr{B}}$$\end{document} B indicate that SR networks perform better than interpolation and they penalize for the artefacts.

Techniques:

Journal: Scientific Data

Article Title: A Real-World Benchmark for Sentinel-2 Multi-Image Super-Resolution

doi: 10.1038/s41597-023-02538-9

Figure Lengend Snippet: The MOS survey outcome showing how often each metric was consistent with the answers (in %), stratified into SR networks trained with PROBA-V and simulated images, and interpolation.

Article Snippet: However, it can be noticed that these qualitative observations are quantitatively reflected only in the LPIPS values, while PSNR and SSIM are slightly worse for HighRes-net and RAMS–overall, both LPIPS and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr{B}}$$\end{document} B indicate that SR networks perform better than interpolation and they penalize for the artefacts.

Techniques:

Journal: Sensors (Basel, Switzerland)

Article Title: Joint Demosaicing and Denoising Based on a Variational Deep Image Prior Neural Network

doi: 10.3390/s20102970

Figure Lengend Snippet: Comparison of the CPSNR, SSIM and FSIMc values for the different CFA patterns used with the proposed method on the Kodak image dataset.

Article Snippet: , Kodak , SSIM-R , 0.8253 , 0.5630 , 0.5372 , 0.5405 , 0.6490 , 0.7927 , 0.8300.

Techniques: Comparison

Journal: Sensors (Basel, Switzerland)

Article Title: Joint Demosaicing and Denoising Based on a Variational Deep Image Prior Neural Network

doi: 10.3390/s20102970

Figure Lengend Snippet: Comparison of the SSIM values among the various demosaicing methods on the Kodak and the McMaster image datasets.

Article Snippet: , Kodak , SSIM-R , 0.8253 , 0.5630 , 0.5372 , 0.5405 , 0.6490 , 0.7927 , 0.8300.

Techniques: Comparison

Journal: Journal of Nuclear Medicine

Article Title: Projection Space Implementation of Deep Learning–Guided Low-Dose Brain PET Imaging Improves Performance over Implementation in Image Space

doi: 10.2967/jnumed.119.239327

Figure Lengend Snippet: Summary of Previous Studies Focusing on LD-to-FD PET Conversion in Brain PET Imaging Using Deep Learning Techniques

Article Snippet: The convolutional network trained in PIS applied simplistic noise reduction, thus leading to blurred, highly smoothed, and slightly biased FD images. table ft1 table-wrap mode="anchored" t5 TABLE 5 caption a7 Study Patients ( n ) Input Network architecture Injected activity (MBq) LD ratio (%) Tracer Scanning time (min) Time from injection to scan (min) Scanner model Evaluation method and metrics ( 16 ) 11 LD PET, MRI (T1) Segmented brain tissues from MRI to build tissue-specific models to predict standard-dose plus iterative refinement strategy 203 ± 12 25 18 F-FDG 12 36 Siemens Biograph mMR Leave-one-out cross-validation, SUV bias ( 8 ) 8 LD PET, MRI (T1, DTI) Sparse representation based on mapping strategy and incremental refinement scheme 203 ± 12 25 18 F-FDG 12 60 Siemens Biograph mMR NMSE, PSNR, contrast recovery, quantification bias ( 19 ) 16 LD PET MRI (T1) Deep auto-context CNN architecture with 4 layers 203 ± 12 25 18 F-FDG 12 60 Siemens Biograph mMR NMSE, PSNR, training and testing time comparison, training loss vs. iteration ( 22 ) 9 LD PET 2.5D U-Net (modified) 370 0.5 18 F-FDG 40 45 GE Signa PET/MRI NRMSE, error maps ( 15 ) 40 LD PET MRI (T1, T2, 3D) 2D U-Net 330 ± 30 1 18 F-florbetaben 20 90–110 GE Signa PET/MRI SSIM, PSNR, RMSE, clinical image quality scoring, Bland–Altman analysis ( 20 ) 16 LD PET 3D conditional GANs 203 ± 12 25 18 F-FDG 12 36 Siemens Biograph mMR SSIM, PSNR, RMSE, quantification bias ( 21 ) 40 LD PET 2.5D GAN 330 ± 30 1 18 F-florbetaben 20 90–111 GE SIGNA PET/MRI SSIM, PSNR, RMSE, clinical image quality scoring, error maps This work in PIS 140 LD PET image 3D U-Net (modified) 205 ± 10 5 18 F-FDG 20 34 Siemens Biograph mCT SSIM, PSNR, RMSE, clinical image quality scoring, Bland–Altman analysis, quantification bias, radiomic features, joint histogram, error maps This work in PSS 140 LD PET sinograms 3D U-Net (modified) 205 ± 10 5 18 F-FDG 20 34 Siemens Biograph mCT SSIM, PSNR, RMSE, clinical image quality scoring, Bland–Altman analysis, quantification bias, radiomic features, joint histogram, error maps Open in a separate window DTI = diffusion tensor imaging; GAN = generative adversarial network; NMSE = normalized mean square error; NRMSE = normalized root mean square error.

Techniques: Imaging, Comparison, Modification