The main advantage that QTVR offers over other three-dimensional visualisation techniques is that it is remarkably quick, cheap and relatively simple to use. After the required images have been captured, an attractive product is very quickly ready for presentation over the Internet. In addition to this, the capturing of landscapes (or objects) as photographs rather than surface models means that a level of detail is available that is simply not the case with three-dimensional modelling packages without the careful rendering processes described above or the integration of photographs into the model. This means that virtually all landscape details and features normally visible from any point in the landscape will be visible in the photographic panorama, depending on original image resolution and QTVR compression levels. The main advantage that this gives QTVR over three-dimensional modelling is that it allows all landscape features, up to and including the actual horizon, to be visible rather than the artificial horizon created by a three-dimensional model.
In terms of interactivity, QTVR does not compare well to a three-dimensional model presented in, for example, VRML. QTVR allows the generation of QTVR scenes which are essentially linked sequences of panoramas. This means that the user is able to navigate around a landscape from node to node via hotspots, but the user cannot explore the landscape freely as there is no underlying landscape model to explore. This can make what looks like a large open landscape feel constrained and restrictive simply because of the nature of the interaction with the landscape representation. It also places a much larger burden of interpretation on the originator of archaeological QTVR panoramas, as the nodes chosen will be assumed to be archaeologically significant (i.e. locations from which the landscape was viewed in the past). An example of this might be a node offering a view of the landscape from a hilltop overlooking a valley containing a Pictish symbol stone, combined with a node at the stone. Here the user may potentially assume that the hilltop viewpoint had some significance in the past, whereas the originator may have chosen this point arbitrarily as simply one with a view of the stone. A better strategy might be a sequence of nodes on possible lines of approach and a node at the stone. Whilst feeling far more restrictive in the range and type of movement it allows in comparison to VRML, QTVR does offer the originator the opportunity consciously to control the viewer, perhaps in a way that echoes control of movement in the past. It is essential here to differentiate between nodes intended to give an overview of the landscape for general orientation purposes and those which the originator thinks might represent either a significant or common viewpoint in the past. This may be done by inserting text directly into the image or by linking the panorama to explanatory text. Despite the additional levels of interactivity offered by VRML, it is perfectly possible to restrict these, and therefore the movements of the user, through a VRML model or landscape in order to convey a particular impression, as described above for QTVR landscapes.
It is only possible to combine a model of an object, such as a stone cross, directly into a QTVR panorama as a two-dimensional image, inserted into the panoramic tile using an image manipulation package, such as Adobe PhotoShop (Fig. 6 a & b). This is similar to the common technique of inserting two-dimensional shots of computer-generated objects into landscape photographs, but without the additional 360 degree panning round the landscape. Dennis Holloway, a New Mexico architect with a strong interest in reconstructing native American settlement sites, has produced many excellent examples of his models inserted into 2D landscape images (Holloway 2000).
Figure 6 a & b: Screen shot of a QTVR panorama with a two-dimensional image of a reconstructed three-dimensional model inserted into it. The results are quite unsatisfactory because although the landscape can be seen for 360 degrees, there is no interaction with the model of the cross (the back face cannot be viewed). The panorama below (Figure 6b), taken from the summit of Dunadd fort, Argyll, used in this example
Inserting models into 2D images or panoramic images has the added disadvantage of requiring any alteration of the landscape, for archaeological purposes, to be done using image manipulation packages such as Adobe PhotoShop. Depending on the complexity of the alterations required this may be inexpensive, but does not compare well with the flexibility regarding lighting, weather conditions and so on offered by three-dimensional modelling, and rendering packages. It would be possible to represent a landscape at sunrise and sunset using QTVR, but two sequences of images would have to be created, taken in the field or 'touched-up' later, to capture these two conditions effectively. In a modelling/rendering package, on the other hand, virtual lighting conditions are easily altered. Similarly, if there is an archaeological requirement for vegetation cover to be altered to match what would be predicted by palaeoenvironmental survey, then modelling and rendering packages provide that flexibility, without the labour-intensive work involved to alter a sequence of images.
© Internet Archaeology
URL: http://intarch.ac.uk/journal/issue10/jeffrey/section4.html
Last updated: Wed May 16 2001