Topographic maps are arguably one of the most information-dense, yet intuitively usable, graphical artifacts produced by mankind. Cartography as science and practice has developed and collected a wealth of design principles and techniques to cope with the problems of high graphical density, especially for the case of label placement. Many of the more sophisticated techniques that take into account figure-ground relationships for lettering have not been fully operationalized until now. We present a novel generic quality evaluation model that allows full automation of refined techniques for improving map feature overlap, visual contrast and layer hierarchy. We present the objective function as a set of metrics corresponding to the design principles and provide exemplary parameterization via the set of experiments on global real-world datasets. The approach designed for labeling of point-like objects and can potentially be applied to linear and areal features. It has a low computational and memory requirement. Furthermore, it is conceivably applicable to annotate any kind of visualization beyond maps. The results of the conducted tests and comparison with a commercial labeling package illustrate the ability to produce highly legible and readable map lettering with our approach. Presented method heeds more cartographic design principles and is computationally less costly compared to commercially available methods.