Deep S2P: Integrating Learning Based Stereo Matching Into the Satellite Stereo Pipeline

TL;DR

This paper integrates modern learning-based stereo matching algorithms into the satellite stereo pipeline, demonstrating improved accuracy and detail in digital surface models from satellite imagery, while highlighting ongoing challenges in natural environments.

Contribution

It adapts and incorporates recent learning-based stereo matchers into the satellite pipeline, enhancing surface model accuracy and detail over classical methods.

Findings

Improved Digital Surface Model accuracy with learning-based methods

Qualitative results show sharper structures and better geometric detail

Performance on challenging natural surfaces remains limited

Abstract

Digital Surface Model generation from satellite imagery is a core task in Earth observation and is commonly addressed using classical stereoscopic matching algorithms in satellite pipelines as in the Satellite Stereo Pipeline (S2P). While recent learning-based stereo matchers achieve state-of-the-art performance on standard benchmarks, their integration into operational satellite pipelines remains challenging due to differences in viewing geometry and disparity assumptions. In this work, we integrate several modern learning-based stereo matchers, including StereoAnywhere, MonSter, Foundation Stereo, and a satellite fine-tuned variant of MonSter, into the Satellite Stereo Pipeline, adapting the rectification stage to enforce compatible disparity polarity and range. We release the corresponding code to enable reproducible use of these methods in large-scale Earth observation workflows. Experiments on satellite imagery show consistent improvements over classical cost-volume-based approaches in terms of Digital Surface Model accuracy, although commonly used metrics such as mean absolute error exhibit saturation effects. Qualitative results reveal substantially improved geometric detail and sharper structures, highlighting the need for evaluation strategies that better reflect perceptual and structural fidelity. At the same time, performance over challenging surface types such as vegetation remains limited across all evaluated models, indicating open challenges for learning-based stereo in natural environments.