A Canonic Framework for Sequences of Images: Similarities between Calibrated and Uncalibrated Case

Invariance has become increasingly important in computer vision. One type of invariance is viewpoint invariance, i.e. properties of viewed objects that are invariant under camera motion. Another important concept is object invariance, i.e. properties of the image plane that only depend on motion and not on the viewed objects. The epipolar constraint is such an example. This paper deals with the generalisation of the epipolar constraint to the problem of analysing sequences of images taken by calibrated or uncalibrated cameras. It is assumed that the correspondences between the points in the different images are known. In a previous paper it is shown how the problem of analysing sequences of images taken by an uncalibrated camera can be simplified by choosing a specific affine basis in each image plane. In this paper we show that this theory has remarkable similarities with the theory for calibrated cameras.