| 摘要 |
<p>An image according to a scalable encoding scheme having enhancement and reference layers is encoded. For encoding the enhancement layer, a block predictor candidate (4.6) is obtained (5.1 in fig. 5) for predicting an enhancement layer coding unit (4.5) and an associated first residual block corresponding to the prediction. A reference layer block predictor (4.12) co-located with the enhancement layer predictor candidate (4.6) is identified (5.2 in fig. 5). A second residual block is determined (5.3 in fig. 5), associated with the reference layer coding unit (4.8) co-located with the enhancement layer coding unit (4.5), using the determined reference layer block predictor (4.12 NOT 4.7). This second (reference layer) residual block does not correspond to the residual used in the usual reference layer predictive coding (which is based on 4.7). A further residual block is determined (5.4 in fig. 5) for the enhancement layer coding unit (4.5) corresponding, at least partly, to the difference between the first and the second residual blocks. Differences between the enhancement layer motion and the co-located reference layer block motion are handled better. The proposed invention constitutes a generalisation of Inter Layer Residual Prediction (ILRP) modes, called GRILP (Generalized Inter Layer Prediction Mode).</p> |
