Image transformations in mountainous terrain and the relationship to surface patterns

Abstract

Image transformations based on the normalized cosine correction (Lambertian model) and the hyperspherical directional cosine transform are computed for SPOT satellite Multispectral Linear Array (MLA) and LANDSAT Multispectral Scanner (MSS) and Thematic Mapper (TM) imagery of four mountainous, subalpine, and boreal terrains. Linear discriminant functions are used to determine the power of the original and transformed data sets in separating land systems, land-cover types, geomorphic classes, and vegetation communities and for interpretation of physical conditions such as vegetation dynamics. These results are compared to those that can be obtained by incorporating topographic variables from digital elevation models directly into the discriminant functions. After transformation, certain classes such as those containing large amounts of deciduous forest vegetation are separated more accurately.