Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learned Image Compression for Earth Observation: Implications for Downstream Segmentation Tasks

Christian Mollière, Iker Cumplido, Marco Zeulner, Lukas Liesenhoff, Matthias Schubert, Julia Gottfriedsen

TL;DR

This work addresses the challenge of transmitting and storing increasing EO data by evaluating task-specific learned image compression against JPEG 2000 across fire, cloud, and building segmentation tasks in multispectral and thermal domains. It uses a differentiable end-to-end framework based on a Discretized Gaussian Mixture Likelihood compressor (CompressAI) and a U‑Net segmentation backbone, comparing single-channel and multi-channel settings. The findings show that learned compression offers substantial gains on multi-channel optical imagery, improving PSNR and segmentation accuracy, while JPEG 2000 remains competitive for small, single-channel thermal tasks; end-to-end optimization provides no clear benefit over standalone optimization. The results guide practical compression strategy recommendations for EO missions, emphasizing spectral richness and the trade-offs between downlink reduction and downstream segmentation quality. Future work should broaden single-channel EO coverage and build larger annotated datasets to improve learned codecs under realistic constraints.

Abstract

The rapid growth of data from satellite-based Earth observation (EO) systems poses significant challenges in data transmission and storage. We evaluate the potential of task-specific learned compression algorithms in this context to reduce data volumes while retaining crucial information. In detail, we compare traditional compression (JPEG 2000) versus a learned compression approach (Discretized Mixed Gaussian Likelihood) on three EO segmentation tasks: Fire, cloud, and building detection. Learned compression notably outperforms JPEG 2000 for large-scale, multi-channel optical imagery in both reconstruction quality (PSNR) and segmentation accuracy. However, traditional codecs remain competitive on smaller, single-channel thermal infrared datasets due to limited data and architectural constraints. Additionally, joint end-to-end optimization of compression and segmentation models does not improve performance over standalone optimization.

Learned Image Compression for Earth Observation: Implications for Downstream Segmentation Tasks

TL;DR

This work addresses the challenge of transmitting and storing increasing EO data by evaluating task-specific learned image compression against JPEG 2000 across fire, cloud, and building segmentation tasks in multispectral and thermal domains. It uses a differentiable end-to-end framework based on a Discretized Gaussian Mixture Likelihood compressor (CompressAI) and a U‑Net segmentation backbone, comparing single-channel and multi-channel settings. The findings show that learned compression offers substantial gains on multi-channel optical imagery, improving PSNR and segmentation accuracy, while JPEG 2000 remains competitive for small, single-channel thermal tasks; end-to-end optimization provides no clear benefit over standalone optimization. The results guide practical compression strategy recommendations for EO missions, emphasizing spectral richness and the trade-offs between downlink reduction and downstream segmentation quality. Future work should broaden single-channel EO coverage and build larger annotated datasets to improve learned codecs under realistic constraints.

Abstract

The rapid growth of data from satellite-based Earth observation (EO) systems poses significant challenges in data transmission and storage. We evaluate the potential of task-specific learned compression algorithms in this context to reduce data volumes while retaining crucial information. In detail, we compare traditional compression (JPEG 2000) versus a learned compression approach (Discretized Mixed Gaussian Likelihood) on three EO segmentation tasks: Fire, cloud, and building detection. Learned compression notably outperforms JPEG 2000 for large-scale, multi-channel optical imagery in both reconstruction quality (PSNR) and segmentation accuracy. However, traditional codecs remain competitive on smaller, single-channel thermal infrared datasets due to limited data and architectural constraints. Additionally, joint end-to-end optimization of compression and segmentation models does not improve performance over standalone optimization.

Paper Structure

This paper contains 22 sections, 1 equation, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Experimental Setup
  3. Datasets
  4. Segmentation Baseline Methods
  5. Compression Algorithms
  6. Optimization Methodology
  7. Results
  8. Segmentation Baseline Performance
  9. Segmentation under Bit Rate Reduction
  10. Standalone Compression Performance
  11. Task-specific Compression Performance
  12. Conclusion
  13. Channel-dependent Efficacy
  14. Implications for Downstream Tasks of EO Missions
  15. Future Directions
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Results of segmentation baseline models.
  • Figure :
  • Figure :