Chapter 4 deals with the extraction and matching of features and the recovery of multiple view relations. A robust technique is presented to automatically relate two views to each other.
Chapter 5 describes how starting from the relation between consecutive images the structure and motion of the whole sequence can be built up. Chapter 6 briefly describes some self-calibration approaches and proposes a practical method to reduce the ambiguity on the structure and motion to metric.
Chapter 7 is concerned with computing correspondences for all the image points. First an algorithm for stereo matching is presented. Then rectification is explained. A general method is proposed which can transform every image pair to standard stereo configuration. Finally, a multi-view approach is presented which allows to obtain denser depth maps and better accuracy.
In Chapter 8 it is explained how the results obtained in the previous chapters can be combined to obtain realistic models of the acquired scenes. At this point a lot of information is available and different types of models can be computed. The chapter describes how to obtain surface models and other visual models. The possibility to augment a video sequence is also presented.