Cai et al. (2025) JSPSR: Joint Spatial Propagation Super-Resolution Networks for Enhancement of Bare-Earth Digital Elevation Models from Global Data

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-10-30
• Authors: Xiandong Cai, Matthew Wilson
• DOI: 10.3390/rs17213591

Research Groups

The provided text does not explicitly list the research groups, labs, or departments involved.

Short Summary

This research introduces the Joint Spatial Propagation Super-Resolution network (JSPSR) to convert global Digital Elevation Models (DEMs) into bare-earth DEMs with enhanced spatial resolution and vertical accuracy. The method significantly outperforms existing techniques, achieving an RMSE of approximately 1.1 meters for 3-meter and 8-meter resolution bare-earth DEMs derived from 30-meter global DEMs.

Objective

• To develop an innovative method to convert global Digital Elevation Models (DEMs) to bare-earth DEMs while simultaneously enhancing their spatial resolution and improving vertical accuracy.

Study Configuration

• Spatial Scale: Global DEMs (specifically Copernicus GLO-30 DEMs) downscaled to 3 m and 8 m spatial resolutions. Experiments conducted in low-relief areas.
• Temporal Scale: Not explicitly mentioned as a temporal study; focuses on static DEM generation.

Methodology and Data

• Models used: Joint Spatial Propagation Super-Resolution network (JSPSR), which integrates Guided Image Filtering (GIF) and Spatial Propagation Network (SPN). Compared against bicubic interpolation and a baseline Single Image Super Resolution (SISR) method.
• Data sources:
  • Copernicus GLO-30 DEMs (30 m spatial resolution) as input.
  • Guidance features extracted from remote sensing images (with or without auxiliary spatial data).
  • Open-access data used to develop a real-world bare-earth DEM Super-Resolution dataset.

Main Results

• JSPSR improved prediction accuracy by 71.74% on Root Mean Squared Error (RMSE) and reconstruction quality by 22.9% on Peak Signal-to-Noise Ratio (PSNR) compared to bicubic interpolated GLO-30 DEMs.
• It achieved 56.03% and 13.8% improvement on RMSE and PSNR, respectively, against a baseline Single Image Super Resolution (SISR) method.
• The overall RMSE for JSPSR-generated DEMs was 1.06 m at 8 m spatial resolution and 1.1 m at 3 m spatial resolution.
• This compares favorably to GLO-30 (3.8 m RMSE), FABDEM (1.8 m RMSE), and FathomDEM (1.3 m RMSE) at either resolution.

Contributions

• Presents an innovative deep learning method (JSPSR) for bare-earth DEM super-resolution, integrating GIF and SPN.
• Demonstrates significant improvements in both spatial resolution and vertical accuracy of bare-earth DEMs compared to state-of-the-art global DEMs and other super-resolution techniques.
• Addresses the critical need for higher-resolution, more accurate bare-earth DEMs by effectively correcting elevation errors introduced by surface features.
• Developed a novel dataset for real-world bare-earth DEM Super-Resolution problems in low-relief areas using open-access data.

Funding

The provided text does not contain information regarding specific funding projects, programs, or reference codes.

Citation

@article{Cai2025JSPSR,
  author = {Cai, Xiandong and Wilson, Matthew},
  title = {JSPSR: Joint Spatial Propagation Super-Resolution Networks for Enhancement of Bare-Earth Digital Elevation Models from Global Data},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17213591},
  url = {https://doi.org/10.3390/rs17213591}
}