Satellite Image Fusion using FDCT for Land Cover Classification

Abstract

Remote sensing is a fast developing field of science involving repetitive collection of data from earth observing satellites. However each satellite system has one or more limitations, giving rise to the need of data collection from multiple sources and their fusion. Landsat 8 collects images in a broad spectrum but at a coarser spatial resolution of 30m. Cartosat-1 collects images at a high spatial resolution of 2.5m but lacks color details. Good visually interpretable images are indispensable for land cover classification. In this paper, the Landsat 8 and Cartosat-1 images are fused by using the Fast Discrete Curvelet Transform (FDCT) method. Supervised classification using the Random Forest (RF) classifier is performed on the Landsat 8 multispectral image and the fused image. The results showed high quality of image fusion based on the entropy, RMSE and CC values obtained for the given dataset. The fusion process also improved the overall accuracy of the land cover classification. © 2021 IEEE.