Image sequence coding using a contour-based method

This paper describes a new technique for reconstructing an image sequence from spatio-temporally represented contour information. The contour information includes geometry and amplitude information. The dependency of contours between frames is determined by their shape, amplitude and temporal difference. Contours are connected if the dependency of the contours between adjacent frames is strong. Connected contour groups are reconstructed in the form of a tree structure from dominant contours using their motion vector to each frame. Simulation results show that a constructed image sequence can be obtained with about 128 kbit/s for color CIF format image sequences at 25 frames/s