Comparative performance analysis between topographic correction models for landsat-8 OLI images

Most images acquired in mountainous areas include topographic effects. Many approaches have been proposed and developed to reduce the topographic effects, which are divided into three main categories: 1) lambertian, 2) non-lambertian and 3) empirical models. The correction performances of these models largely depend on land cover types and wavelengths. However, the quantitative comparison between the models has not been carried out. Our objective of this study is to compare the performances of lambertian, non-lambertian and empirical topographic correction models in different wavelengths. For this study, we acquired two Landsat-8 Operational Land Imager (OLI) (bands 2-6) images. And a Shuttle Radar Topography Mission digital elevation model (SRTM DEM) is also used for the comparison of the models. The performance test of the models is performed by using the Hellinger distance between sun-facing and sun-shaded areas. The result shows that the Minnaert correction model is best for the visible and short wave infrared (SWIR) bands. The mean Hellinger distances of the corrected images are 0.01, 0.10, 0.09 and 0.25 in blue, green, red and SWIR bands, respectively, while those of the original images are 0.90, 0.93, 0.91 and 0.94, respectively. However, the best model for the near infrared (NIR) band is the statistical-empirical correction model. The mean Hellinger distance is improved from 0.96 to 0.23. This result demonstrates that the Minnaert correction model is best for the topographic correction of the visible and SWIR images and the statistical-empirical correction model is best for the topographic correction of the NIR band.