Is there any particular paper you could refer to this being mentioned? This may help a bit for others.
The crossing number of an arbitrary graph is NP-hard to solve in general. But, there do exists bounds for some classes of graphs (I'd suspect the algorithm you are talking about draws from one of the constructions used to count them(?)). Are you interested in the crossing number or the rectilinear crossing number? Most proofs bounding such a number usually are constructive. For example, here's a paper on the rectilinear crossing number of K_n (the construction is of course, algorithmic):