| 摘要 |
A multi-dimensionally layered encoding and decoding system having reduced the bandwidth requirements for a channel that carries an image stream encoded by the system. A source image stream is partitioned multi-dimensionally into a lower resolution layer and higher resolution layer image stream. The lower resolution image stream is encoded and decoded in accordance with certain accepted standards such as MPEG-2, MPEG-4 and JPEG. In the encoding of the lower layer image stream, a number of indicator signals are derived that indicate where a compression loss and resolution loss occurs. Additionally, the lower layer image stream is non-linearly processed to create an improved image stream and non-linear indicator for use in the upper-layer encoder. The upper-layer encoder uses the non-linear processed image stream and a decoded, reconstructed upper-layer encoded stream to derive the motion vectors and prediction frames in a motion compensation process. The upper-layer encoder uses the nonlinear processed image stream and indicators as well as motion compensation prediction frames to form a number of difference image streams which are evaluated for their information content. One of the difference image streams having the lowest information content or statistical variance and acceptable quality is selected, encoded and sent as the upper layer encoded image stream. The upper layer encoded image stream and the lower layer encoded stream are combined and sent over the communications channel. A decoder that receives the combined stream separates out the upper and lower encoded streams, decodes the lower stream, reproduces the non-linear processed stream and determines the which difference image stream was sent in the upper-layer encoded stream. Using the non-linear processed stream and the determined difference image stream along with any needed prediction frames, the upper-layer stream is reconstructed.
|
