YES - PSD should be applicable. In image processing the term "spatial frequency" is used. Although being somehow abstract (usually no dimension other than 'pixels' is used as images are randomly scalable in themselves), the transformations for PSD apply well.

Regards

Yes...  in the sense that a power/energy density may exist in the Fourier domain of some function of something other than time...  For example, if we are talking about a function of space, then the Fourier domain is often called "wavenumber" and we can have a power/energy density as a function of wavenumber.  To the extent that you are willing to call a wavenumber domain a "spectrum" then a PSD can be measured in this domain.  Put another way, if you can generate an autocorrelation function across something other than time, then the Fourier Transform of this is a Power/Energy density.

Armin, U. Dreher -

Thank you very much for your answers.   I admit that I'm having a hard time conceptualizing FFT other than it is a method for analyzing a complex waveform and representing it a spectrum of sine waves.  PSD is also conceptually difficult compared to the typical aggregate texture characterizations which have a more tactile, physical nature.

Thanks again,

Mike

Yes - the PSD is an excellent tool to characterize spatial heterogeneity in an image as a function of length scale (1/wavenumber).  It can help you detect the characteristic length-scales of your image, or at the very least the ratio of small-scale structure to large-scale structure.

The Fourier Transform (FT) is a reversible mapping of any function of an independent variable x to a function of an independent variable 1/x. Accordingly, you can use the FT to treat any series (function) of any variable  you like, provided that it is periodic or stationary (its expectation does not change with time). The PSD is merely the amplitude squared of the transformed function in the domain 1/x. So basically, yes