Up: graphics resources

Generating Images

One of the most important steps in a visualization project is to choose the right kind of software for the task. There are several distinct genres of image generating software, each with its own strengths and weaknesses.

2D Structured Graphics

Structured graphics programs keep track of the objects drawn, so that they can be moved and edited later. Typically, a structured graphics program will enable you to draw lines, polygons, ellipses, place text and so on. They do not allow you to "air brush" over the whole scene, since this process is a "pixel" operation, where the color of individual pixels changes in many objects. Conequently, structured graphics tend to be better suited to illustrations and diagram s, as opposed to more "artistic" or photorealistic images. Another useful feature of structured graphics is that sizes and colors tend to be independent of the underlying hardware.

Adobe Illustrator -- Illustrator is a powerful, high end package. It runs only on the NeXTs at the Center; find it in the /LocalApps directory. Illustrator also interfaces well with Adobe Photoshop, which is a pixel based, "raster" drawing program, so between the two, it is possible to produce very high quality graphics. To learn about Illustrator, consult the manuals in the Center library.
xfig -- xfig is a relatively straight forward, bare bones drawing program. It allows only 8 colors, but it outputs PostScript directly, integrates very well with TeX, and it is widely available. This is a good choice for a simple diagram from a mathematical paper in TeX. The man page contains detailed information.
gnuplot -- gnuplot is a graphing program. That is all it does, but it does that relatively well. It can make 2D and 3D plots, either from data files, or constructed out of elementary functions. There are lots of style options. For more information, consult its man page, or the online documentation, which is probably more useful.

Raster Graphics

Raster graphics programs are "paint programs". You cannot go back and select an object and edit it; instead, you change the colors of the pixels on the screen as you go, and the computer only keeps track of the image. Thus, one can "air brush" and apply "effects" in a way one cannot in a vector graphics drawing program. This allows the user to create beautiful and highly refined images, but like a painting, they cannot be easily "edited" later. Thus, raster graphics are best for things like logos and cover artwork -- high quality, but one time graphics. Another thing to watch out for with raster graphics is that they tend to depend on the underlying hardware; colors and sizes will vary from machine to machine.

Adobe Photoshop -- Photoshop is essentially the only raster graphics editor the Center has, and there is only one copy. Ask the staff on which Mac it is currently residing. Photoshop is the graphic arts industry standard, and is a very powerful program. Among other things, it works very well for "touching up" illustrations created with Adobe Illustrator. For more information, consult the manuals in the Center library.
IconBuilder -- A quicker, simpler graphics editor is the Next IconBuilder.app in /NextDeveloper/Apps. It provide roughly "MacPaint" level editing. It only works with .tiff files.

Mathematical Objects

A large proportion of people at the Geometry Center will naturally be concerned with making pictures of objects with mathematical descriptions. This enterprise takes place at two levels. On the one hand, there are occasions when the main objective is simply getting some picture to look at, in order to better understand an object. On the other hand, people also need to produce high quality images to communicate a visual concept.

Typically, one will use specialized mathematical software to produce some graphical representation of an object. Then, if a high quality image is desired, the user will use the graphical description of the object produced in the first phase as input to a 3D modelling and rendering program where lighting, texture mapping, shadows and so forth can be added. For example, one might use Mathematica to compute the vertices of a dodecahedron, and then use Geomview to find a good viewpoint and add lights and shading. If necessary, one could then use Geomview to write a Renderman file, which could then be used to render the dodecahedron in mahogany, through a haze.

There are many ways of producing a first pass graphical representation of a mathematical object. However, four packages have proved particularly useful in the recent experience at the Center. Basic instructions for making mathematical illustrations with them are online:

3D Modeling and Rendering

The generation of high quality 3D images is complicated and labor intensive. In terms of ease of use, Geomview is the clear pick for generating a 3D scene, however Geomview doesn't support texture mapping yet, or a variety of other high end features. However, Geomview can output a scene description in Renderman (.rib) format. By editing the RIB file, one can add texture mapping and so on. Most frames from the Geometry Center videos "Not Knot" and "Outside In" were produced with Renderman. Renderman does not do ray tracing however, so if things like reflections and shadows are important, one must use something like Rayshade or Softimage. Softimage is state-of-the-art, boasting the production credits for Jurassic Park among other things. However, the learning curve for Softimage is immense, so the need for ray tracing should be compelling before undertaking a Softimage project. Similarly, using Rayshade requires learning another scene description language, not to mention producing the actual description.

Geomview -- Geomview is the correct choice for most mathematical 3D modeling. Scenes are described by OOGL files (Object Oriented Graphics Language) which one can create directly, or via the applications noted above. One of Geomview's most outstanding features is the ability to manipulate a scene in real time, and interactively set many rendering attributes. This gives the user the ability to experiment with a scene to find the combination of rendering options and viewing parameters that most effectively conveys a visual concept. Moreover, Geomview can output a scene description in RIB format for further processing by Renderman.
To get started with Geomview, consult the online documentation or the printed manual in the Center library.
Renderman -- Renderman produces high quality still images of scenes described in RIB format. There is a library of C routines one can use to produce a scene description in RIB format, but often one works directly with the RIB file. Although Renderman is not interactive, it is possible to do sophisticated shading and texture mapping.
Softimage -- Starting up Softimage is like climbing into the cockpit of a spaceship, except Softimage has more enigmatic buttons. However, even if it is complex, it is nonetheless an interactive 3D modeling environment, complete with sophisticated tools for building up a scene description as well as rendering it.
Softimage is only licensed on abel, nonabel, gauss and riemann. You will want to start by perusing the documentation in the Center Library.
Rayshade -- Rayshade is a public domain ray tracing program. One writes a scene description file, specifying lighting and rendering options, as well as the objects in the scene. Objects must be describe in terms of constructive geometry, i.e. constructed out of triangles, cylinders, cones, spheres, etc. To get started, consult the printed documentation in the Center library.

Up: graphics resources

The Geometry Center Home Page

Comments to: webmaster@geom.umn.edu
Created: Fri Sep 8 11:39:00 1995 --- Last modified: Jun 18 1996