The atmosphere absorbs light very strongly in certain well-defined wavelength intervals (absorption bands). This causes image bands within those intervals to be relatively dark. The flat field correction technique is used to reduce this effect. This does not result in a reflectance map, but it does simplify the visualization and analysis of the data.
In contrast to the empirical line correction technique, the flat field correction technique does not require ground- or laboratory-measured data as input. Instead, it involves extracting a spectrum (or the mean of several similar spectra) from the image itself. This spectrum would be selected such that it is relatively featureless, or "flat," except for the absorption band features. It would correspond to a material in the scene that has relatively uniform reflectance over broad wavelength ranges. Call this spectrum the reference spectrum. The correction transformation for a band simply consists simply of a gain that is the reciprocal of the reference spectrum value for that band. The application of the transformation corresponds to mostly "dividing out" the absorption band features.
The FTLOC program addressed the problem of finding a flat reference spectrum. It generates an image bitmap that indicates the image locations where the image spectra are best approximated by a polynomial function of wavelength, of user-specified order.
In practice, the average of multiple spectra from spatially neighboring FTLOC-computed locations that represent the same scene material should be used as the reference spectrum, in order to reduce the effect of noise present in the individual spectra.
Once a reference spectrum is derived from the FTLOC-computed locations then it must be converted into a radiometric transformation using SP2RT. This transformation is stored in the image file, and can be applied "on the fly" when the image data are later processed or analyzed. Alternatively, APPLRT may be used to apply the transformation in creating a new, transformed image.
© PCI Geomatics Enterprises, Inc.®, 2026. All rights reserved.