Web Pages

Siggraph 1999 Course Notes on Non-Photorealistic Animation



Craig Reynolds NPR Page

University of Utah's Gooch and Gooch Page

Brown University Research Group Page

New York University Research Group Page

Bibliography of Non-Photorealistic Rendering

Background on NPR

The intent of both Non-Photorealistic Rendering (NPR) and Photorealistic Rendering is to visually communicate information. The Photorealistic Rendering community creates images by simulating the physics of light in mathematically modeled scenes. The goal of Photorealistic Rendering is to create images that are indistinguishable from photographs of real world scenes. In contrast, NPR is a newer and in many ways broader field whose community possesses an assortment of image creation goals. NPR images are created using a variety of methods from the simulation of traditional artistic media to completely ad hoc. The only unifying principle in NPR is that all NPR images are created in order to stimulate the human visual system.

One of the most important skills for an artist is to learn is choosing the correct medium for a given subject. Artists make these choices guided by considerations such as aesthetic appeal and the effectiveness of the medium in communicating the required visual message. The directors of feature films integrate NPR methods with traditional film media to produce effects never before possible. Nearly every three-dimensional rendering package now has a cartoon shading plug-in that empowers nonartists with the ability to create images in ways only professional artists were once able. NPR gives the graphics community freedom to choose media in addition to a camera for creating images.

A common assumption in the graphics community is that NPR involves simulating natural artistic media. This assumption is not surprising, because the first research in the field focused on reproducing traditional art forms, such as pen and ink, watercolor, and oil on canvas. Technology development in any field first seeks to imitate the previous mode of working. Thus, NPR seems to be following the usual development scheme. NPR currently embraces a wider scope of research. For example, a recent trend in NPR research is interactive NPR techniques. Current work focuses on the detection and rendering of feature lines to communicate shape. Silhouettes, surface and texture boundary lines, as well as creases are important for communicating the shape of an object.

The goal of generating computer images that are indistinguishable from photographs is essential for a host of applications including design, marketing, and the entertainment industry. In many applications, however, a nonphotorealistic image has advantages over a photorealistic image. NPR images omit extraneous detail, focus attention on relevant features, clarify, simplify, and disambiguate shape, and show hidden parts.

The control of detail in an image for purposes of communication is becoming the hallmark of NPR images. Often this control of image detail is combined with stylization to evoke the perception of complexity in an image without an explicit representation. NPR images also provide a more natural vehicle for conveying information at different levels of abstraction and detail. Some additional occasions when an NPR image has an advantage are listed below.

The early adoption and subsequent interest in photorealistic rendering by the graphics community is most likely due to the ``mission statement'' of photorealistic rendering: "Create an image that is indistinguishable from a photograph." This mission statement gives photorealistic rendering a visual "Turing test", and an easily defined metric for a successful image. NPR does not have a single mission statement. Instead, researchers are pursuing a number of image creation goals. The goals of NPR include simulating traditional artistic media, understanding the human visual system, communicating effectively with low bandwidth, abstracting images, enhancing learning, and improving user interaction.

In the computer graphics community, rendering is the process by which a virtual scene is converted into an image. Photorealistic rendering has come to mean images depicting what is real based on physical simulations. While physical phenomenon are objective and relatively predictable, photorealism is inherently costly due to the vast quantity of data that must be specified and processed. The main difficulty in striving for photorealism is that the physical world is complex. To render a computer-modeled scene realistically, the scene must approximate its real-world counterpart to a high degree of accuracy, resulting in extremely complex geometry. Therefore, the realistic depiction of complex objects always involves a series of trade offs based on the image creator's time and computational budget.

Images created using a photorealistic rendering process can be objectively measured or visually compared to a photograph of the physical process being simulated. In NPR the goal of the computer generated images is believability. A believable image is one that effectively communicates the intent of the image's creator. Believability is simpler to achieve than photorealism, because a believable model needs only to include the details representative of the intent behind the image.