Research in image synthesis has mainly concentrated on realistic images, which is of course very important for a wide variety of applications, such as AV production, simulation and digital mockups.
On the other hand, for a number of other domains, realism is not even necessarily desirable. For example in architecture, archaeology, or education, a more artistic or "expressive" representation is often more appropriate, since it allows the user to focus on specifics aspects of a scene or an image, rather than be overwhelmed by a multitude of realistic, but possibly irrelevant details. This is particularly important when realism is not, or cannot be the central aspect of the activity under consideration. For example when presenting everyday life in an archaeological context, an exact representation is either unknown or impossible to implement.

It is evident that the manual work of an artist to create such representations is often long and difficult. Since 1994, researchers in image synthesis have developed methods for automatic rendering in different styles, often called "non-photorealistic" rendering, based on 3D models (Winkenbach 94), or as a 2D post-processing in image space (Deussen 00a, Ostromoukhov 99).
 
 

In the figure above we show examples of the work developed by the research group at the University of Washington, led by D. Salesin on non-photorealistic rendering, who are the major pioneers in this domain. Notice that the same 3D model was used to generate the two drawings. This work inspired a number of other papers which simulate different drawing, illustration or painting styles in CG (Winkenbach 94, Ostromoukhov 99 etc.). Most of these methods require significant processing time, and thus cannot be used in an interactive context.

More recently, several methods have been developed for interactive NPR. These methods concentrate on specific styles or used tailored rendering methods (Freudenburg 01), such as the replacement of OpenGL calls at a very low level in the library itself (Mohr 01). There have also been some first attempts at using NPR for virtual environments (Klein 01).

Despite these developments, NPR has not been used extensively in the context of virtual environments and in particular immersive systems such as CAVEs or RealityCenters. We have identified two major problems that explain this:
 

•  Due to the novelty of the NPR work, users and potential applications in non-photorealistic virtual environments have note necessarily been identified. This may be explained in part by the fact that the central issues in VE research, namely "Presence" and Immersion have often been associated with "being believable" thus realistic.

• Significant technological difficulties remain for NPR rendering in a VE context, for example concerning stereo rendering for certain algorithms, and display speed in a number of others.

In this research project we address both issues.

Our consortium consists of participants who are experts in computer graphics (REVES, iMAGIS/ARTIS), archaeologists and historians (ERGA, ENS), architects (ARIA), and experts in usage of VE environments for the presentation of cultural heritage, and notably in an educational context (FHW/VR). Our collaboration will permit the definition of application scenarios for virtual visits of archaeological sites and in particular will allow us to determine when a non-photorealistic representation is desirable in the context of an immersive virtual environment or walkthrough. Constant and close interaction between archaeologists, historians, architects and VE/cultural heritage experts will be the key to success of this project.

We will concentrate on the application related to the virtual visit of Argos, which is a project already initiated by iMAGIS and ERGA, and possibly work on a second project through collaboration of REVES/ENS.

• Our main technical goals are the following: develop interactive display algorithms for ancient buildings, for example comprising columns and walls, in the context of Argos, using pen-and-ink and watercolor rendering, or in a style adapted to the specific usage or user group targeted. We will also develop methods for virtual population of these sites for such visits. Sound simulation could also be added to add a stronger sense of immersion.
• We will examine problems related to the usage of LOD and appropriate drawing styles for different hypotheses. We also wish to examine the problem on uncertainty of archaeological hypotheses and the appropriate style used for its depiction.
• The visualisation and representation of different periods in time will be investigated as a second step, allowing the user to navigate in the temporal dimension.

We will define the application in close collaboration with the users, and we hope that the novelty of our approaches will lead to the presentation of one or two demonstrators at the end of the project on immersive virtual reality systems, such as the RealityCenter in Grenoble or the workbench at Sophia-Antipolis. The demonstrator may be presentation at the celebration for the 50 years of the Argos site, and if the results are satisfactory, it is possible that they be presented to the general public in Athens.