Supporting Automated Pen and Ink Style Surface Illustration with B-Spline Models

Article excerpt


In the 19th century, William Henry Holmes, Grove Karl Gilbert, and William Morris Davis perfected the construction of pen and ink style representations of physiographic forms and processes (Fernlund 2000). Rooted in the tradition of perspective landscape illustration expressed in the drawings and paintings of da Vinci and the later woodcuts of Murer and other 16th century artists, their illustrations are characterized by a restrained application of strokes, each contributing a critical aspect of landform structure or process. Examples of their art are characterized both by renderings of view-dependent features, such as the silhouette of a hill against the sky, or view-independent aspects of surface morphology, such as hydrologic ridges or valleys. Later artists such as Armin Lobeck (1939; 1958), Erwin Raisz (1948) and Eduard Imhof (2007), produced a number of excellent works detailing both the utility and construction of pen and ink style landscape representations. Lobeck argues that such images achieve an effective portrayal of natural surface forms primarily through the artist's selection of only those elements to which the viewer's attention should be directed (Lobeck 1958, pp. 1-2). Imhof (2007, p. 45) echoes Lobeck and provides a succinct comparison of landscape photographs (and, by extension, shaded surfaces) to line drawings, stating:

"Views of nature and their photographic reproduction contain no lines. The drawn line is a human invention, a useful abstraction or fiction. Forms and their spatial relationships can be clarified by lines, the essential can be emphasized and the nonessential can be subdued. A good sketch is simpler, more expressive and aesthetically more satisfying than a photograph ... It should never be forgotten that sketching means leaving things out!"

Lesage and Visvalingam (2002) note that cartographic sketches facilitate visual processing of static images by focusing attention on significant features. Santella and DeCarlo (2004) show this experimentally through eye tracking experiments, noting that subjects extract image information by looking at fewer locations in non-photorealistic rendering (NPR) abstractions than they do in literal photographic reproductions. Although significant progress in creating NPR landscape renderings has been made over the past decade, most attempts continue to fall short of the quality exhibited by hand-drawn examples. Some of this disparity can be attributed to the use of planar-faceted digital elevation models--built either from regular grid cell or triangulated irregular network structures--as the underlying basis for surface analysis. Planar surface tessellations derived from triangulated samples rarely provide sufficiently smooth analytical surfaces for landscape drawing. Imhof (2007, p. 41) notes that classically trained terrain illustrators use surfaces of ideal geometric solids (cylinders, cones, pyramids, and other forms) as models for landscape representations. By providing a surface of generalization, such forms help the artist to filter out local elements of surface roughness that would otherwise obscure dominant surface features. This paper presents a methodology, its implementation, and results for feature extraction and pen and ink style rendering from B-spline polynomial surface models. Polynomial surface representations are not, of course, new to automated cartography. Maxwell and Turpin (1968) propose the use of polynomial models for topographic surface representation for civil engineering projects. Although attractive with respect to the minimal data storage capabilities of the 1960s, such models require untenable execution times when applied to the modeling and rendering of fine-resolution surface processes. Fortunately, as Imhof and Lobeck stress, fine surface detail is often unwanted in pen and ink style landscape representations. Furthermore, 3rd degree polynomial surface patches can be 'stitched' together while maintaining continuity at joins. …