| 摘要 |
A method for editing a coded residual-error signal is presented which, in particular, is suitable for transmitting image sequences via a channel having a limited transmission capacity. By comparing video images succeeding one another in time, the similarity of these images is utilised by the fact that, for part-image areas, motor vectors are determined which describe the movement of the part-image areas. By a comparison, it can be determined at the transmitter end whether the reconstructed image sequence obtained on the basis of the information on the displacement of the part-image areas sufficiently accurately corresponds to the original sequence. For part-image areas of the image in which this is not the case, the transmission of picture- element-dependent information, the so-called residual-error signal, is required in addition to the transmission of the motion vectors. The method according to the invention takes into consideration, on the one hand, the special structure and statistics of the residual- error signal to be transmitted. On the other hand, characteristics of the human optical sense are utilised for obtaining as low as possible a data rate for describing the residual-error signal. For this purpose, the residual-error signal is subjected to an adaptive coarse quantisation in dependence on a previous classification. This classification is effected in accordance with the effectiveness of a preceding motion compensation. |
