DBPapers
DOI: 10.5593/SGEM2014/B23/S10.004

ASSESSING THE PERFORMANCE OF RELATIVE RADIOMETRIC NORMALIZATION METHODS FOR SOME VEGETATION INDICES

I. Vorovencii, C. C. Teresneu, M. M. Vasilescu
Wednesday 1 October 2014 by Libadmin2014

References: 14th International Multidisciplinary Scientific GeoConference SGEM 2014, www.sgem.org, SGEM2014 Conference Proceedings, ISBN 978-619-7105-12-4 / ISSN 1314-2704, June 19-25, 2014, Book 2, Vol. 3, 27-34 pp

ABSTRACT
Using the time series of satellite images in studies on vegetation indices involves absolute or relative radiometric normalization (RRN). In this study we used two Landsat Thematic Mapper 5 images acquired in 2007 and 2011, the first representing the subject image and the second the reference image. Five methods of RRN have been applied to the subject image acquired in 2007 for evaluating their performance in relation with vegetation indices. These methods include simple regression (SR), pseudo-invariant features (PIF), no-change set determined from scattergrams (NC), histogram matching (HM), and radiometric control set (RCS). They were compared in terms of their statistical robustness and capability to improve visual image quality. The focus was on the manner in which each different RRN method influences the results of extracting information about vegetation indices. It was found that each vegetation index after RRN differs in comparison with the same vegetation index before RRN. We also identified the factors which affect the performance of RRN in the case of the analyzed vegetation indices.

Keywords: relative radiometric normalization, vegetation indices, simple regression, pseudo-invariant features, histogram matching, satellite images.