A PCA and Mesh Adaptation-Based Format for High Compression of Earth Observation Optical Data With Applications in Agriculture

Abstract

Earth Observation optical data are critical for agriculture, supporting tasks like vegetation health monitoring, crop classification, and land use analysis. However, the large size of multispectral and hyperspectral datasets poses challenges for storage, transmission, and processing, particularly in precision farming and resource-limited contexts. This work presents the H${\hspace{0.17em}}^2$-PCA-AT (Hilbert and Huffman-encoded Principal Component Analysis-Adaptive Triangular) format, a novel lossy compression framework that combines PCA for spectral reduction with anisotropic mesh adaptation for spatial compression. Adaptive triangular meshes capture image features with fewer elements with respect to a standard pixel grid, while efficient encoding with Hilbert curves and Huffman coding ensures compact storage. Numerical evaluations of data reconstruction, vegetation index computation, and land cover classification demonstrate the effectiveness of the H${\hspace{0.17em}}^2$-PCA-AT format, achieving superior compression compared to JPEG while preserving essential agricultural insights.