Surface nets
Isosurface extraction that places one vertex per cell near the surface and joins neighbouring cells into a mesh — a dual method, simpler and smoother than marching cubes.
Where marching cubes emits triangles inside each cell from a lookup table, surface nets is a dual method: every cell the surface crosses gets a single vertex, placed at the average of the edge crossings (and optionally relaxed toward its neighbours), and quads are formed by joining the vertices of the four cells sharing an active edge. The mesh has fewer, more uniform polygons and no thin slivers.
Sarah Frisken Gibson introduced "naive surface nets" in 1998. Dual contouring extends the idea with Hermite data and a quadric solve so hard edges and corners survive, at greater cost; plain surface nets stays cheap and rounds everything off slightly. Both run just as happily on a voxel field.