I will review some old ideas about the duality between matrix models and Liouville theory coupled to a  (p,q) minimal model. In this approach t'Hooft diagrams of the matrix model are thought of as a discretization of a 2d surface. Fine-tuning of the coupling constant leads to divergence of the perturbative expansion. In this limit dense t'Hooft diagrams dominate and become a good approximation of a smooth 2d surface. 
Reading material:
hep-th/9306153;  hep-th/9304011; Seiberg, "Notes on quantum Liouville theory and quantum gravity"; Lecture notes by Zamolodchikov brothers http://qft.itp.ac.ru/ZZ.pdf