A Novel Data Compression Techniques for Remote Sensing Data Mining
Résumé
In this article we propose a parameter-free method for Remote Sensing (RS) image databases Data Mining (DM). DM of RS images requires methodologies robust to the diversity of context found in such large datasets, as well as methodologies with low computational costs and low memory requirements. The methodology that we propose is based on the Normalized Compression Distance (NCD) over lossless compressed data. Normalized Compression Distance is a measure of similarity between two data files using the compression factor as an approximation to the Kolmogorov complexity. This approach allows to directly compare information from two images using the lossless compressed original files, and avoiding the feature extraction/selection process commonly used in pattern recognition techniques. This shortcut makes the proposed methodology suitable for DM applications in RS. We provided a classification experiment with hyperspectral data exemplarizing our methodology and comparing it with common methodologies found on the literature.
Fichier principal
A._Roman_Gonzalez_-_A_Novel_Data_Compression_Techniques_for_Remote_Sensing_Data_Mining_-_2011.pdf (236.11 Ko)
Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...