Two dimensional filters for digital processing of images: Design and implementation techniques

Two design methods of two dimensional digital filters are presented. Finite impulse response filters which make possible linear phase frequency responses are emphasized. The design methods considered are windowing and transformations. In the windowing methods, a time limited function is sought, the Fourier transform of which would be a best approximation to a given frequency limited function. An approximate solution for the one dimensional case is the Kaiser window: a two dimensional Kaiser window is proposed. In the transformation method, a change of variables transforms a one dimensional frequency response into a two dimensional one. Finally, implementation of filters and their application is discussed for the case of LANDSAT images.