Digital images are composed of Spatial Frequencies which describe "change" with respect to position in space. For more information please see: http://en.wikipedia.org/wiki/Spatial_frequency
In fact, the two-dimensional Fourier Transform can decompose a given image to its spatial frequencies which means that it can decompose an image to a set of "sinusoidal plates". Think of these plates like egg-cartons (http://www.asia.ru/images/img/alibaba/img/product/11/24/61/11246140.jpg) of different density (egg positions per unit of space). That is, sinusoids in 2D. For more information, please see: http://cns-alumni.bu.edu/~slehar/fourier/fourier.html and http://homepages.inf.ed.ac.uk/rbf/HIPR2/fourier.htm
As a rule of thumb, keep in mind that small features (i.e. object details) occupy high spatial frequencies while large features (i.e. object form) occupy low spatial frequencies.
A comprehensive example of this would be high-pass filtering of an image which would preserve the edges of represented objects (an edge being a very sudden change in light reflectance across a direction) but completely lose all information about the rough form of the image. The opposite of this would be low-pass filtering which would completely wipe out the details (sharp transitions of reflectance across a direction) but preserve the very slow transitions of reflectance within an image. For more information on shaping the spatial content of an image please see: http://en.wikipedia.org/wiki/Kernel_%28image_processing%29
Hope this helps.