This kind of makes sense from a mathematical point of view, but how would this look implementation-wise, in a scenario where you need to render a polygon scene? The article states that box filters are "the simplest form of filtering", but it sounds quite non-trivial for that use case.