2.0 Some current approaches to three-dimensional representation

There are two technologies that are currently in general use for creating three-dimensional representations of objects in a landscape context. The first includes three-dimensional modelling packages such as 3D Studio Max, the product being presented for the Internet as VRML, Quicktime Movies or .AVI animations. The second is Quick Time Virtual Reality (QTVR) which is designed specifically for the Internet. These technologies take very different approaches to creating the illusion of three dimensionality on a two-dimensional screen. Although QTVR itself could be legitimately described as a '3D modelling package', here it is differentiated from other such packages by its reliance on two-dimensional images, normally photographs, as opposed to three-dimensional information as described below.

Three-Dimensional Modelling packages were originally developed as part of the computer-aided design (CAD) process. Given the large amount of data they typically handle and the processing requirements of that data, they are generally very sophisticated software packages. They create their models by holding information on an object in three dimensions. This information consists of an object's position in three-dimensional space as well as its shape in three dimensions. Generally, although not always, this is done by assigning three-dimensional co-ordinates (XYZ) to a number of points on the surface of the object, with the surface itself being represented by planar polygons defined by these points (Fig. 1 a & b). Clearly most real world objects, such as sculpted stones and landscapes, are not actually constructed in this way; that is, they are not an amalgam of straight-sided geometric shapes. Despite this, they can be adequately represented in this fashion as long as there are enough points and the points are well distributed. However, the larger the number of points, the larger the files required to store this information and the higher the level of computational complexity required to manipulate them. In practice this means that a level of accuracy is decided on by the model originator, either as a sampling strategy during the data capture phase of constructing a model, or later, depending on hardware constraints or the final presentation medium of the model.

Govan Sarcophagus point cloud Govan Sarcophagus plain surfaces
Figure 1 a & b: These images are derived from a photogrammetrically generated model of the Govan Sarcophagus, an ornate stone sarcophagus, possibly influenced by Pictish sculpture (Spearman 1994, 42). Fig 1a (on the left) shows a point cloud, where the points are on the surface of the object to be modelled and defined relative to each other in three-dimensional space. Fig 1b (on the right) shows polygons bounded by the points in order to define the surface of the object

Quick Time Virtual Reality (QTVR) approaches the representation of three-dimensional objects and environments in a completely different way from the three-dimensional modelling packages described above. QTVR technology was designed by Apple specifically for the Internet and treats objects and environments in a slightly different way from each other. For environments, a series of images, normally photographs, taken from a single location are stitched together to form a 360 degree panorama. This panoramic image is then conceptually projected onto the inside of a cylinder, and the user is then able to view the image through a window situated at the centre of the cylinder, referred to as a node, panning round the cylinder and also zooming in and out. This is known as pseudo-immersive virtual reality and the technique has exploded in use in the last year, primarily on commercial web sites but increasingly on archaeological web sites where landscape context and visualisation are important (e.g. McElearney 1999). For an example of a QTVR panorama see Figure 6b.

QTVR treats discrete objects in yet another way, effectively the opposite of the environmental panoramas described above. Here the concept is of the object sitting in the centre of a globe and a series of images is taken of the object from around the inside surface of the globe. The images are then presented in an interactively defined sequence that has the effect of animating the movement from image to image, making the object appear to rotate in the viewing window. In order for this rotation to appear smooth the horizontal and vertical angular separation of the original images must be carefully controlled. This is ordinarily done using a turntable or a custom built 'object rig' (Fig. 2). These cannot be used with large immovable objects, such as a cross-shaft still in situ. In these cases other methods for controlling the camera positions must be used, such as using an EDM to position the camera accurately, though in practice this is extremely difficult.

Kaidan C-60 object rig
Figure 2: An object rig in operation. The camera is attached to the moveable, curved, arm and can photograph the object, in this case an iMac, from any angle with exact precision. This object rig is manufactured by Kaidan who kindly gave permission for the use of this image

Obviously the techniques of modelling and QTVR are not intended to be interchangeable, although it is possible to present animations generated in three-dimensional modelling packages using the Quick Time Player media plug-in. Each technology was developed with differing objectives and therefore has differing capabilities. What follows is a slightly more detailed look at the advantages and disadvantages of each technique, specifically in relation to presenting sculptured stones.


© Internet Archaeology URL:
Last updated: Wed May 16 2001