S. Saran et al., Extraction and refinement of landuse/landcover information from satellite digital data using fast Fourier transformation in spatial frequency domain, CURRENT SCI, 78(1), 2000, pp. 87-90
Applicability of fast Fourier transformation (FFT), which deals with a co-o
rdinate space known as spatial frequency domain, has an advantage in the pr
ocess of feature extraction for different landuse/landcover classes over th
e single supervised classification. The study deals with the extraction of
low frequency component from the power or amplitude spectrum by applying in
verse Fourier transform to convert back into spatial domain in the form of
image. The derived Fourier transformed and filtered image is partially devo
id of radiometric error which generally prevails in the remotely sensed dat
a due to the atmospheric effect, This can be further minimized, whereby enh
ancing the texture of the image, The classified map extracted from the FFT
image provides a better estimation off the landcover classes than the super
vised classification alone. This method has applicability for mapping of na
tural resources and monitoring them.