We describe progress in completely automatically recovering 3D scene structure together with 3D camera positions from a sequence of images acquired by an unknown camera undergoing unknown movement.

The main departure from previous structure from motion strategies is that processing is {em not} sequential. Instead a hierarchical approach is employed building from image triplets and associated trifocal tensors. This is advantageous both in obtaining correspondences and also in optimally distributing error over the sequence.

The major step forward is that closed sequences can now be dealt with easily. That is, sequences where part of a scene is revisited at a later stage in the sequence. Such sequences contain additional constraints, compared to open sequences, from which the reconstruction can now benefit.

The computed cameras and structure are the backbone of a system to build texture mapped graphical models directly from image sequences.