Date of Original Version
Abstract or Description
Many physical phenomena have complex structure in both space and time. To systematically understand these phenomena from images we need representations that unify the treatment of space and time. We create such a representation, the space-time map, for characterizing contour evolutions. Many types of information are computed and stored as facets of the map, registered to a space-time manifold generated by the evolution. We demonstrate our representation on the example of Drosophila embryogenesis in optical section. Changes in embryo shape are reflected as changes in the dye distribution along the deforming vitelline membrane contour. We extract a series of contours and create a two-dimensional space-time map. We track intensity on this map to obtain a velocity field. We extract space-time ridges and significant motions on this map, and use them along with prior knowledge to recognize the significant features and events of embryogenesis.
Proceedings of the Workshop on Biomedical Image Analysis, 144-153.