Processing remote sensing images using the 2-D FFT—Noise reduction and other applications

Abstract

With the development of faster and less expensive computers, it is now practical to employ algorithms for processing remote sensing images that were only feasible on large mainframe computers a short time ago. The two‐dimensional (2-D) fast Fourier transform (FFT) is a powerful means for removing noise because it can be used to design and implement efficient filters based on the observed spatial or frequency patterns of the noise. Illustrations of transforms of image points and lines in various configurations and periodic patterns are used as a tutorial to identify and reduce a variety of noise patterns of increasing complexity from an assortment of spacecraft and aircraft systems. A general strategy for filtering developed from this study involves: filtering only derived products (not original images), initially removing noise evident in the transform, and applying a “minimum” filter to reduce residual noise. The 2-D FFT is also applied to mosaicking, enlargement, registration, and extraction of albedo and slope information.