Lithologic Discrimination by Fourier Analysis of Image Texture

Drainage density is influenced by the erosion resistance of the underlying rocks. Ideal circumstances for lithologic discrimination by measurement of drainage network parameters are: 1) Geomorphic maturity 2) Relatively homogeneous rock structure 3) Slope as the only source of gray-tone modulation (e.g., in jungle) An example of such a terrain would be a crystalline shield area in the wet tropics. Our reconnaissance study involves the use of digital topographic tapes and aircraft and spacecraft (Seasat-A) radar data. The digital topographic data are first processed by SHADY 2, an algorithm that creates a scene illuminated at a selectable geometry according to a cosine scattering law. SHADY 2 has been modified to offer radar-like scattering behavior and to generate shadows. The Fourier transforms of the images are computed and filtered with ten annular bandpass filters. The inverse transforms of the bandpass-filtered data are generated to produce images containing spatial features having a restricted range of dimensions. Spatial regions having high activity in a given frequency band will have a large variance in these images, so we generate a variance picture whose DN\u27S are a measure of the local variance of the input image. These variance pictures (~10) are used as inputs in a supervised classifier. Frequency slice pictures of the Klamath Mountains of Northern California show: 1) Only weak dependence on illumination geometry 2) Distinct clumping of high-activity areas 3) Differences in the location of the high-activity areas as a function of frequenc