摘要 |
A moving picture sequence is divided into key frames, which are coded by a conventional method, and non-key frames, which are coded to obtain error-correcting information. The encoder and decoder independently generate predicted images for each non-key frame. The encoder compares its predicted image with the original non-key frame to estimate how much error-correcting information the decoder will need to decode the non-key frame by correcting prediction errors in the decoder's predicted image, and sends the decoder that amount of error-correcting information together with correlation information indicating how closely the predicted image and the original non-key frame are correlated. Before decoding the non-key frame, the decoder modifies the correlation information to compensate for differences in the prediction methods employed by the encoder and decoder. The modification enables the decoding process to converge faster.
|