A geostatistical approach for selecting the highest quality MODIS daily images

Pesquer L., Domingo C., Pons X. (2013) A geostatistical approach for selecting the highest quality MODIS daily images. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 7887 LNCS: 608-615.
Link
Doi: 10.1007/978-3-642-38628-2_72

Abstract:

The aim of this work was to develop a new methodology for automatic selection of the highest quality MODIS daily images, MOD09GA Surface Reflectance product. The methodology developed here complements the quality assessment of MODIS products with a geostatistical analysis of spatial pattern images based on variogram tools. The resulting selection is formed by 26 high-quality images (from an initial dataset of 365) from throughout 2007. Most images with geometric distortion problems, such as the bow-tie effect, were rejected. The automatic selection was validated by comparing it to manual selection, which showed that it achieved an overall accuracy of 71.4%. © 2013 Springer-Verlag.

Read more

Spatial pattern alterations from JPEG2000 lossy compression of remote sensing images: Massive variogram analysis in high performance computing

Pesquer L., Pons X., Cortes A., Serral I. (2013) Spatial pattern alterations from JPEG2000 lossy compression of remote sensing images: Massive variogram analysis in high performance computing. Journal of Applied Remote Sensing. 7: 0-0.
Link
Doi: 10.1117/1.JRS.7.073595

Abstract:

We evaluate the implications of JPEG2000 lossy compression of remote sensing images for spatial analytical purposes. The main issue is to identify which cases and conditions in geostatistical studies are suitable for using lossy compressed images. For these purposes, an extensive test using Landsat, compact airborne spectrographic imager (CASI), and moderate resolution imaging spectroradiometer (MODIS) image series has been analyzed, through applying and comparing two-dimensional and three-dimensional (spectral and time domains) compression methods with a wide range of compression ratios for several dates, different landscape regions, and spectral bands. Due to the massive test bed and consequently to the high time consuming executions, a parallel solution was specifically developed. Variogram analyses showed that all the compression ratios maintain the variogram shapes, but high compression ratios (>20:1) degrade the spatial patterns of the remote sensing images. These alterations are lower for the three-dimensional compression method, which was a considerable improvement (25%) on the two-dimensional method for large three-dimensional series (CASI, MODIS). However, the two methods behave similarly in the Landsat case. Finally, the parallel solution in a distributed environment demonstrates that high performance computing offers a suitable scientific platform for highly demanding time execution applications, such as geostatistical analyses of remote sensing images. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License.

Read more