The application of the covariance matrix statistical method for removing atmospheric effects from satellite remotely sensed data intended for environmental applications

Abstract

The Covariance Matrix Method (CMM) uses the statistical relationship between all the selected bands of a satellite sensor simultaneously, rather than one at a time as in the regression method. It examines the set of variances and covariance between all band pairs in the image data and CMM provides an average pixel correction for a specified part of a satellite image. It is necessary to know a priori a value for the atmospheric path radiance on one spectral band. From this, CMM enables the estimation of the atmospheric path radiances in all the other bands. Dark pixels must be present in the CMM technique. Indeed, the authors suggest an improved CMM atmospheric correction algorithm. This methodology has been presented as an improved revised version of the CMM atmospheric approach. The authors provide a critical assessment of the suitability of the CMM atmospheric correction using Landsat TM image data of an area consisting low reflectance targets that have been used for several environmental monitoring applications. The proposed improved method produces retrieved surface reflectance within the range of the ground measurements