There are two well-explored methods for construction of a modelfrom range imagery: surface adjacency graph and mesh computation. Thescale of a method is basically the fineness of the edges that thealgorithm uses. Proper setting of the scale can make a crucialdifference in the accuracy or quality of the output model. Dynamic scalealgorithms use algorithmic variation of the scale with feedback from theresult to help find the best scale, or for more robust reconstruction bycombining the results from different scales. This report discusses thegeneral approach well (with 28 references), develops criteria forevaluating results of model construction, and presents experimentalresults on a large number of scenes from four sources.