Fourier-Modulated Implicit Neural Representation for Multispectral Satellite Image Compression

2025-06-02

AI Summary

To address the challenges of compressing and reconstructing high-dimensional, large-volume multispectral satellite imagery with cross-band spatial resolution heterogeneity, this paper proposes ImpliSat, an end-to-end implicit neural representation (INR) compression framework. Methodologically, it introduces a novel Fourier modulation mechanism to dynamically adapt to spectral and spatial characteristics across bands, and establishes a unified implicit representation framework—enabling, for the first time, joint modeling and compression of multi-resolution heterogeneous bands. The approach integrates INRs, coordinate encoding, frequency-domain modulation, and multi-scale feature alignment. Evaluated on Sentinel-2 and Landsat datasets, ImpliSat achieves a mean PSNR of 32.7 dB at a bitrate仅为 38% that of JPEG2000, significantly outperforming state-of-the-art INR-based and conventional compression methods.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Data Mining & Knowledge Management: Data Compression

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Multispectral satellite images play a vital role in agriculture, fisheries, and environmental monitoring. However, their high dimensionality, large data volumes, and diverse spatial resolutions across multiple channels pose significant challenges for data compression and analysis. This paper presents ImpliSat, a unified framework specifically designed to address these challenges through efficient compression and reconstruction of multispectral satellite data. ImpliSat leverages Implicit Neural Representations (INR) to model satellite images as continuous functions over coordinate space, capturing fine spatial details across varying spatial resolutions. Furthermore, we introduce a Fourier modulation algorithm that dynamically adjusts to the spectral and spatial characteristics of each band, ensuring optimal compression while preserving critical image details.

Problem

• Compress high-dimensional multispectral satellite images efficiently
• Handle diverse spatial resolutions across multiple channels
• Preserve critical image details during compression

Innovation

• Uses Implicit Neural Representations for compression
• Models images as continuous spatial functions
• Applies Fourier modulation for spectral adaptation