This paper introduces a novel power spectral density model for a class of images. This model is parametric with a variable number of parameters, and is easily defined in the correlation domain through the product of basic building blocks. By taking advantage of certain naturally occurring structural characteristics common to image power spectra, the result is capable of providing spectral detail that previous approaches could not. Consideration of these structural characteristics also allows accurate spectral estimates to be obtained even in the presence of some corruption and from low resolution data. That is, for typical images the model and estimation procedure provide a degree of robustness against some amounts of noise, distortion, and aliasing. The model and a numerical algorithm for estimating its parameters are presented along with some discussion and an example.