On the Internet, the transmission rates and the quality of the transmission paths vary with the user's connection environment. Thus, to provide image transmission services on the Internet, bitstream scaling techniques are essential to convert to any transmission rate in real time. Already proposed scaling methods include methods that reduce the number of coding bits for the DCT coefficients and methods that discard coded frames not referenced by other frames. However, the degrees of freedom in scaling in previous methods are inadequate to provide image transmission services on the Internet. In this paper, we propose a bitstream scaling method that discards the reference frames to limit the computational costs. In the proposed method, instead of a search for motion vectors, new motion vectors are estimated from the original motion vectors and the reference relationship in motion-compensated prediction is altered. The degrees of freedom in scaling can be expanded while limiting the computational costs. The estimated motion vectors are compared to the motion vectors detected from the decoded frame to verify the accuracy of the motion vector estimation method.
ASJC Scopus subject areas