Using PCA transformation to remove the tenuous cloudiness effect in multispectral satellite sensor images

Abstract

An algorithm using no external data is proposed for removing the inhomogeneous effect of thin cloudiness and other aerosols on multispectral satellite sensor images such as Landsat Enhanced Thematic Mapper (ETM) images. The method consists of a series of digital processing operations and is based on principal component analysis (PCA). The goal is to generate, for every original band, a new band whose digital number (DN) values are related only to the atmospheric intensity effect. An example is shown and some limitations are discussed.