Nils Einecke, Julian Eggert, Sven Hellbach, Edgar Körner, "Walking the manifold without falling off", Proceedings of the 14th International Conference on Neual Information Processing (ICONIP), pp. 653-662, 2007.

Having a good description of an object’s appearance is crucial for good object tracking. However, modeling the whole appearance of an object is difficult because of the high dimensional and nonlinear character of the appearance. To tackle the first problem we apply nonlinear dimensionality reduction approaches on multiple views of an object in order to extract the appearance manifold of the object and to embed it into a lower dimensional space. The change of the appearance of the object over time then corresponds to a walk on the manifold, with view prediction reducing to a prediction of the next step on the manifold. An inherent problem here is to constrain the prediction to the embedded manifold. In this paper, we show an approach towards solving this problem by applying a special mapping which guarantees that low dimensional points are mapped only to high dimensional points lying on the appearance manifold.