At the Lunch ‘n Learn seminar on February 14, Computer Science Professor Adam Finkelstein presented “Modeling by Drawing.”

Computer graphics has progressed marked from its foundations in 1963 when Ivan Sutherland created the first graphical input and output devices using nothing but simple electronics and an oscilloscope. Today, the graphical rendering of a face can be fantastic or, in some cases, a spectacular film disaster when we become so enchanted with technology that we fail to tell a good story.

The tools have progressed to the point that home computers can rival the capabilities of production facilities at movie theatres. But the studios still have dozens of artists working on images, often trying to make sure that the images like Yoda don’t distract from real characters or get in the way of the plot. There’s simply no way at home to replicate that kind of effort or result. Imagine creating a spectacular 3-D image and then trying to give it motion.

Professor Finkelstein presented images that show that there is a fundamental difference between photorealism, essentially efforts to create graphical images that appear like photographs, and non photorealistic rendering [NPR] in which images are more artful representations with greater degrees of abstraction and stylization. Those new to the field are often surprised that we still have a lot to learn about how to create NPR imagery. But the rewards are already clear. As David Byrne put it, “the crooked line sometimes has more soul than the straight one.”

For many uses, it’s often more appropriate to have an image that looks as it were drawn. Abstract images, for example, often provide an opportunity to engage the imagination. But what details ought to be included or left out? NPR often tries to emphasize what’s important, and omit unneeded details, to represent the image abstractly… in essence to simulate what many abstract artists have done for centuries, from the use of special brush strokes and the simulation of watercolor, to permitting viewers to control the image, making it move or animating a character smoothly within the image.

Dr. Finkelstein explained that today’s tools often rely upon contour lines but that researchers have found a new similar kind of line called “suggestive contours”. Contours are lines that denote transitions, obvious boundaries or silhouettes in any given line of sight. As the lines of site change, so too would the contours. Modern tools are clever in predicting mathematically the lines that fine artists draw instinctively.

There are also clever techniques to draw the viewer’s eye to a portion of an image by increasing the density of detail of that portion of the image.

It’s now possible to compose a model first by drawing a rudimentary sketch. The tools can automatically provide three dimensionality and then style, perhaps a wood-like surface or watercolor texture.

You might try out Teddy, a freeform 3d drawing program in 1999 by Takeo Igarashi, an associate professor at the University of Tokyo. Teddy can take a simple hand drawn shape and animate it. It’s very easy to use but provides limited functionality.

At the other end of the spectrum is Maya, a more flexible but much harder-to-use package from Alias for producing computer animations. Finkelstein wondered whether we might soon have a more flexible tool without sacrificing too much ease of use.

One example is Sketchup. Google has recently acquired and made it available for free. There is a professional version, but Dr. Finkelstein demonstrated that even the free version provides a useful modeling tool and impressive stylistic flexibility.

The presentation and a podcast is available.

Posted by Lorene Lavora