Normal Mapping Technology: Parthenon Case Study
Vassiliou, Nikos, Psistakis, Joseph, Zafiris, Paris, Papatheodorou, Theodore S., Journal of Digital Information Management
This paper describes our recent work on using normal mapping for the real-time visualization of 3D reconstructions. Using this technology, now supported on low-cost graphics cards, we can radically reduce the high geometric complexity of reconstructed or laser-scanned 3D models by adding surface detail through normal map images. These images are generated by using the geometric detail of a high polygon model in a low resolution equivalent version. Thus, per pixel lighting can be achieved on higher frame rates, giving the bandwidth for other lighting effects, such as shadow volumes, with less depth and stencil artifacts, leading to superior image quality. We demonstrate and benchmark this technology on a high resolution 3D reconstruction of the Parthenon of Athens, using OpenGL, showcasing the benefits of the normal mapping technology on similar cultural heritage projects.
Keywords: Normal mapping, bump mapping, 3D reconstruction. real-time visualization
3D graphics are becoming more and more important to the preservation of our cultural heritage, since they are a very useful tool for the visualization of archaeological data. In this paper, we will try to show how normal mapping technology can be used in order to increase the visual quality of real-time 3D cultural visualizations, by sacrificing geometric detail and accuracy, and explain its importance in educational applications.
The first step of a 3D cultural visualization is the digitization of the content. This is performed either by automatic content acquisition or by 3D modelling. The first method can be applied on existing objects and can be done through 3D digitizers, such laser scanners, or photogrammetry. These techniques usually provide models of very complex point clouds that capture the very detail of the physical objects. For objects that are ruined or do not exist, such as ancient monuments, manual 3D modelling (reconstruction) is required. Using architectural measurements, drawings and text references, artists try to create an approximation of the original objects using 3D modelling tools, such as AutoCAD, 3DSMax, Lightwave etc. Accurate 3D reconstructions will also provide us with very complex models.
Now that we have acquired our digital content, we must visualize it. For static visualization, such as still images and videos, the digital content can be rendered as is, providing superior accuracy and image quality. In this paper we study the real-time case, in which we try to present our 3D content through real-time CDROM applications, caves, immersive theatres, web etc.
The most common problem with real-time rendering is the hardware limitations on processing thousands of polygons. Although graphic cards have advanced a lot, there are still some issues on the memory transfer. Transferring millions of geometric data from the main memory to the graphic card suffers from bandwidth bottleneck. In addition, using very complex models may cause z-artifacts, due to z-buffer precision problems, and more advanced lighting techniques, such as shadow volumes, are dependent on the polygon count.
One of the most common solutions is to use low resolution models, but we loose too much detail through this technique. An improvement would be to define several versions of varying quality on our 3D models and implement level of detail . According to the distance from the camera we select and draw the appropriate level. The disadvantage of this method is that a very disturbing "popping" effect is generated when we switch between models. At last, other techniques incorporate the use of impostors, 2D textured quads which always face the camera, but we loose all the depth quality of our visualized models.
What we propose here, is the use of the normal mapping technique on culture heritage projects, a technology that we borrow from today's state of the art computer games. …