Given a picture of N1 x N2 dimensions, its pixel values are f [p,q],
the DF tarnsform of this picture is D[m,n], (1) shows how to compute,
(2) is the inverse transformation. D[m,n] is a complex number.
(3) is the base of a computational speedup: first we compute the sum of
the inner sum for each row, then the outer sum for each column.

Further explanation is in every Image processing and Signal processing book.