The problem is that this kind of overview model is too coarse to be used for realistic walk-throughs around the site or for looking at specific monuments. Therefore it is necessary to integrate more detailed models into this overview model. This can be done by taking additional image sequences for all the interesting areas on the site. These are used to generate reconstructions of the site at different scales, going from a global reconstruction of the whole site to a detailed reconstruction for every monument.

These reconstructions thus naturally fill in the different levels of details which should be provided for optimal rendering. In Figure 9.20 an integrated reconstruction containing reconstructions at three different scales can be seen.

At this point the integration was done by interactively positioning the local reconstructions in the global 3D model. This is a cumbersome procedure since the 7 degrees of freedom of the similarity ambiguity have to be taken into account. Researchers are working on methods to automate this. Two different approaches are possible. The first approach is based on matching features which are based on both photometric and geometric properties, the second on minimizing a global alignment measure. A combination of both approaches will probably yield the best results.