Autocartography

Project proposal for Mike Cammarano

In Cartographic Relief Presentation, Imhof describes many techniques developed in the Swiss map making tradition, and illustrates how these principles can be applied to create aesthetically pleasing maps. Many contemporary applications involve generating maps on demand in response to requests for directions or other GIS data. Temporal and economic constraints clearly preclude the introduction of a skilled human cartographer into the interactive map-making pipeline. However, current systems for automatically generating maps by computer generally produce quite ugly results compared to the subtle shading and intricate linework of handcrafted maps.

I propose to incorporate some of Imhof's principles in an automated mapping system. There are a number of subtopics that this could involve; I will list several that I find interesting, roughly ordered by how significantly I feel they contribute to the aesthetic appeal of the maps:

Oblique hill shading
Shading the heightfield given by elevation data seems straightforward, but Imhof notes several subtleties. A global choice of lighting conditions may be inadequate if it fails to differentiate local features. Imhof's maps can vary the lighting conditions and contrast across the map to better portray local topography. Tone mapping techniques might be applicable here.
Hypsometric tints
Many hypsometric color schemes used for computer-generated maps are fluorescent nightmares of psychedelia. See Ray Sterner's California Map from the landform atlas at Johns Hopkins space remote sensing lab. Using color schemes like those in Imhof's book should bring great relief.
Linear elements
Obviously want antialiasing, and should choose colors and line weights effectively.
Generalization / level-of-detail
Large-scale maps must simplify underlying elevation data to meet aesthetic criteria about smoothness both of shading and of linear elements like contours.
Placement of elevation contours
Choosing the number and spacing of contour lines can presumably be cast as an optimization problem.
Text labels
Lots to do here. Automatic placement of labels is a pretty open problem. Also, Imhof doesn't appear to mention it explicitly, but high contrast detail behind text should be muted or removed to improve the legibility of the label.
Hachuring
Unclear how effectively techniques involving fine hatching and texture will map to relatively low-res screen display.
At a minimum, I would like to produce a system that can take raw elevation data and produce an appealing shaded relief illustration.

A reasonable extension would be to address road map and route planning depictions. The MapBlast example shown here could benefit from more careful compositing of the text layer above the roads, for example. I believe that improved line-drawing and compositing techniques would significantly clean up ugly artifacts and poor text legibility of the road maps. Furthermore, introducing subtle background texture and shading in the style of the Swiss relief maps could begin to make these road maps more pleasing to the eye.