| 摘要 |
A server (10) is arranged to stream one of a plurality of encoded streams of video data to a client (40, 50, 60). Each of the plurality of data streams is an independent representation of a common data source encoded at a different resolution from the other of the plurality of data streams. Each stream is encoded as a series of pictures predictively encoded in dependence on the previous pictures in the stream. Linking pictures are generated, each being a picture of the second stream predictively encoded in dependence on a previous picture in the first stream, whereby a receiver receiving the first stream may then receive a linking picture followed by pictures of the second stream. Each linking picture is coded by
(a) quantising the picture of the second stream that is to be coded;
(b) quantising the picture in the first stream that is to be used as prediction;
(c) subtracting the quantised prediction from the quantised picture to be coded to obtain a prediction error signal that is transmitted without further quantisation. Pictures interspersed at predetermined points in the second data stream contemporary with the linking pictures are access pictures to which are applied a quantisation that matches that applied to the corresponding linking picture, such that a receiver can reconstruct a linking picture identically to the simultaneous access picture. The method includes testing an access picture to identify any area of the picture for which the prediction is a good representation such that no data needs to be transmitted and, in respect of such area, setting said quantisations in the access picture and in the associated linking picture to a finer quantisation. |
