Reducing the effects of space-varying, wavelength-dependent scattering in multispectral imagery

Abstract-- A new method for reducing space-varying, wavelength-dependent scattering in multispectral imagery caused by smoke and haze is described. It is intended for use in situations where atmospheric scattering affects the shorter wavelengths and varies in space. The method converts an image in which space-varying scattering is present into an image where the scattering has been equalized over the entire image so that previously developed techniques for removing constant scattering effects can be used. BACKGROUND Assume that the atmosphere can be modeled as a horizontally homogeneous medium, the earth is a Lambertian reflector, and atmospheric properties which vary exponentially with altitude can be assumed to be constant over the scene [ l]. When the contribution of the ambient illumination (skylight) can be neglected and the scattering is constant over the image, the brightness r is the radiance of the surface, a represents the effects of atmospheric transmittance, sensor gain, and other multiplicative factors, and b represents the contribution of the path radiance. The brightness values at two or more pixels in the image together with measurements of the surface radiance at the corresponding locations on the ground can be used to solve for a and b. Alternatively, when ground truth is not available several methods have been developed to estimate the path radiance component [2-41. When the scattering is not constant a technique developed by Lavreau [5] which subtracts a portion of the fourth tasseled cap feature [6] from each band can be used. The method effectively removes space-varying haze but causes the corrected bands to become more correlated thus altering the spectral balance of the data. METHODOLOGY When the effects of scattering are not constant over the image, the image formation model becomes