Author(s)

G.A. SKIANIS, D. VAIOPOULOS, K. NIKOLAKOPOULOS

Abstract

Vegetation indices are usually defined and evaluated empirically, according to their performance on images of areas with certain vegetation types and targets of interest. In this paper we propose a probabilistic approach to the problem of assessing the efficiency of a vegetation index, more precisely the transformed vegetation index (TVI), in its two versions (TVIa and TVIb). A proper distribution is introduced in order to describe the histograms of the red and near infrared channels. Then, the mathematical expressions for the distribution of the TVIa and TVIb values are derived, according to theorems of statistics. The study of the behavior of this distribution shows that the standard deviation of TVIa is bigger than that of the more often employed normalized differences vegetation index (NDVI). This theoretical prediction is verified using satellite images of various regions in Greece and in the Mediterranean Sea. The signal to noise ratio of TVIa and TVIb images is also studied and it is shown that this ratio is bigger than that of NDVI, if the brightness value in the near infrared channel is considerably bigger than that of the red channel. The general conclusion is that TVIa produces images with a good contrast and TVIb presents a good signal to noise ratio over areas with a rich vegetation cover.

Keywords

NDVI, signal to noise ratio, standard deviation, TVI, TVIa, TVIb