Contour-based hybrid displacement estimation for image sequence compression

A contour-based hybrid scheme for video sequence displacement estimation is described. It is a combination of a contour-based and a multiframe gradient-based algorithm. By using three successive image frames, the proposed system can select reliable motion-related-frame differences, extract moving edge tokens, and ignore noise. The contour-based estimation method can provide a reasonable initial displacement estimate for each moving edge neighborhood. As only one motion vector per edge neighborhood needs to be transmitted and the receiver can find the edge and region information by itself, this approach can reduce the overhead bits to be transmitted. To complete a low-bit-rate video coding system, some coding algorithms need to be applied to transmit the motion-compensated frame differences.<>