By background independence I presume you mean coordinate independence. Yet you seem to have tied "duality" into this in some unspecified way. Background independence is a property of coordinate systems such that a choice of scale is arbitrary, i.e., has no real physical meaning. Basically just a bookkeeping device for comparing relative scales, which are physically meaningful. Though background independent formalisms exist which do not specify scale. Any choice of scale can be added after the fact by choosing the units that define the constants.
General Relativity is fully background independent. In fact GR provides the prototype of background independence. The situation is somewhat more subtle in QM. Hilbert space is characterized by an abstract linear space, such that the inner product of vectors has definable meaning. Then by adding Lorentz covariance these vectorial components can be transformed in equivalent components under boost. Yet it is more generally accepted that Hilbert space representations are themselves a mere bookkeeping device lack real physical significance in itself. Though some people beg to differ under highly variable ontological justifications.
Now in what way "duality" or complementarity is related to coordinate independence, or what way you may have intended to imply this meaning, I do not know. As long is there is a relative scale you might define this ratio as some sort of dependence, such as in relativity where relative scales absolutes when expressed as ratios. Yet in the sense "background" in meant in terms of background independence it refers to the background coordinate system independent of any relative relation to things in that background. As such background independence cannot strictly be recovered by weakening the notion of a background to merely mean some distinct observable in that background.
<opinion>
In my opinion ontologies fall into this category of neither being right or wrong, such that you can choose ontologies as freely as you choose coordinate systems. The symmetries of nature only entail that whatever ontology you choose must impose distinct constraints on the ontologies associated with rest of the system in question. Hence there is no more one a true ontology than there is one true coordinate system. However, to date, nobody has been able to construct a completely valid ontological model based on realism. Especially the kind of realism that Newton's critiques invoked to object to his ad hoc fields.
