Using Geostatistics for Data Comprehension and Reconstruction of Remote Imagery. Final Report on Phase 2.
Abstract
This report contains a summary of two of the three earlier interim reports together with the most recent results on image restoration. In addition to the analysis of the remotely sensed data we investigated Paul Krauser's climatic data (Part 1). The climatic variables for the eastern USA were estimated by both auto- and co-kriging, and the results compared with the validation data. In general co-kriging produced more precise estimates than auto kriging. Part 2 contains a brief evaluation of Genstat and Isatis for geostatistical analysis. Part 3 describes the results of an optimal sampling analysis, of kriging from the SPOT information at Fort Benning and of a kriging analysis to filter the imagery so that the short- and long-range structures present could be distinguished more clearly. Part 4 describes the most recent work on data compression and reconstruction. Part of the original area was chosen for this analysis and then sampled at different intensities. The sample data were used for estimation with the variogram model by kriging and conditional Gaussian sequential simulation. The results show that with a good variogram model it is possible to restore the data well from 25% of the original values. Kriging provides the more precise local values, but simulation retains the original degree of variation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 31, 1996
- Accession Number
- ADA325083
Entities
People
- M. A. Oliver
Organizations
- University of Reading