The objective of this study was to use a simplified method based on NDVI classes for the sharpening of the Landsat 5 - TM surface temperature images (Ts) obtained during the years of 2005 and 2006. Thus, three sharpening models, based on the linear regression method, were proposed and compared. The relative and the root mean square errors obtained through the suggested models were of 0.37% and 1.38 ºC, respectively, while the original model presented root mean square error of 1.32 ºC. It was verified that the errors obtained with the accomplished calibrations did not significantly influence in the average values of the thermal images and the results contributed substantially to the improvement of their spatial resolution. The sharpening allowed the precise identification of the targets and features undetectable at the original spatial resolution. This evidences that the simplified method, suggested in this study, allows an accurate sharpening more easily applicable than the original model.
Sharpening; temperature; vegetation indices; NDVI
