Efficient Continuous DEM Reconstruction under Limited Target-Resolution Supervision

2026-09-24

AI Summary

This study addresses the challenge of efficiently reconstructing continuous digital elevation models under limited-resolution supervision. We propose SCOPE, a framework that learns continuous terrain via implicit neural representations by decoupling coefficient prediction from mesh construction. Its core mechanism involves learning a low-resolution reusable coefficient field and integrating local Fourier residuals through geometry-guided fusion, enabling accurate reconstruction beyond the supervised scale. Extensive evaluations demonstrate that SCOPE leads across six primary metrics and reduces errors on unseen scales by 12%. Notably, achieving nine-fold denser reconstructions incurs only a 2% increase in computational overhead. Furthermore, it significantly outperforms baselines in oceanic domains in terms of RMSE, realizing high-fidelity continuous terrain modeling with minimal computational cost.

Abstract

High-resolution digital elevation models (DEMs) support Earth observation applications, but paired training references are often available only at coarser output resolutions. Reconstructing finer terrain grids therefore requires both effective transfer beyond the supervised scale and control of dense-query computation. To address this problem, SCOPE learns a continuous terrain representation from coarser-resolution pairs. It predicts a latent coefficient field on the low-resolution grid and reuses local Fourier residual functions through basis evaluation and geometry-guided ensemble fusion. This separates high-dimensional coefficient prediction from output-grid construction. Experiments on geographically distributed land--ocean samples assess supervised reconstruction, unseen-scale inference, cross-domain generalization, and theoretical computation. SCOPE leads the compared methods across six metrics in the main supervised-scale evaluation. At an unseen factor three times the training factor, land reconstruction reduces RMSE and MAE by approximately 12\% relative to bicubic interpolation, with errors close to target-scale fine-tuning. Ninefold output density increases counted multiply--accumulate operations by only about 2\%. Frozen-model validation on held-out external marine regions reduces RMSE relative to the DEM-specific implicit baseline EBCF-CDEM by approximately 19\% under self-downsampling and 2\% with cross-product inputs, while also yielding lower RMSE than LIIF-MS in both settings. These results demonstrate the value of reusable coefficient fields for accurate reconstruction beyond the supervised resolution with low incremental arithmetic cost.

Problem

• Digital Elevation Model
• Continuous Reconstruction
• Super-Resolution
• Limited Supervision
• Computational Efficiency

Innovation

• Continuous DEM Reconstruction
• Latent Coefficient Field
• Fourier Residual Functions
• Scale Generalization
• Computational Efficiency