Multi-view coding with exploitation of renderable portions
Rendered portions of a renderable portion of a first view of a multi-view signal are introduced into a prediction loop of a multi-view encoder to form a reference signal for the block-based prediction of the encoder's view predictor may improve the prediction and thereby increasing the coding efficiency. The introduction may be performed by completely inserting the renderable portion into the prediction loop to form new reference pictures with their own reference picture indices in addition to reference pictures obtained by the reconstructed version of the multi-view signal of the block-based prediction. Alternatively, the rendered portion may be introduced into the prediction loop completely, but with replacing the normal prediction reference signal, i.e., the normally reconstructed signal as obtained by block-based prediction. Alternatively, it may be signaled within the data stream which portions of the renderable portion are used to replace respective portions in the normally reconstructed signal.
1. A decoder for reconstructing a multi-view signal representing a video, the decoder comprising:
an extractor configured for extracting, from a data stream, using a processor, a residual signal;
a view predictor configured for performing, using the processor, a block-based prediction of the multi-view signal based on a previously reconstructed portion of the multi-view signal and a previously rendered portion determined from the previously reconstructed portion, to obtain a prediction signal;
a reconstructor configured for reconstructing, using the processor, the multi-view signal, using the prediction signal and the residual signal, to obtain a reconstructed version of the multi-view signal;
a renderer configured for at least partially rendering, using the processor, a renderable portion of a first view of the multi-view signal, from the reconstructed version, to obtain an actually rendered portion; and
a replacer configured to replace a co-located portion of the renderable portion with the actually rendered portion.
2. The decoder of claim 1 , wherein the extractor is further configured for extracting, from the data stream, first position information indicative of the actually rendered portion of the first view.
3. The decoder of claim 1 , wherein the extractor is further configured for extracting, from the data stream, second position information indicative of the renderable portion.
4. The decoder of claim 1 , wherein the renderer is configured for rendering the renderable portion to generate the actually rendered portion based on depth/disparity map of a scene associated with the first view.
5. The decoder of claim 1 , wherein the reconstructor is configured for reconstructing, based on the prediction signal and the residual signal, a reconstructed reference signal to be used by the view predictor for block-based prediction.
6. The decoder of claim 1 , wherein the view predictor is configured for performing block-based disparity compensated prediction to obtain a reference signal for a renderable portion of a third view.
7. The decoder of claim 1 , wherein the view predictor is configured to, in performing the block-based prediction, predict at least one first block within the renderable portion using first disparity vector data, and at least one second block within non-renderable portion using second disparity vector data, wherein the second disparity vector data is predicted from the first disparity vector data.
8. The decoder of claim 1 , wherein non-renderable portion of the first view is disjoint to the renderable portion of the first view.
9. An encoder for encoding a multi-view signal representing a video into a data stream, comprising:
a determinator configured for determining, using a processor, a renderable portion of a first view of the multi-view signal, by view synthesis based on a second view of the multi-view signal to generate a rendered portion of the first view;
a view predictor configured for performing, using the processor, a block-based prediction of the multi-view signal based on a previously encoded portion of the multi-view signal and the rendered portion of the first view, to obtain a prediction signal;
a residual generator configured for determining, using the processor, a residual signal between the prediction signal and the multi-view signal;
a parameter predictor configured for predicting, using the processor, a second set of motion prediction parameters based on a first set of motion prediction parameters; and
a data stream generator configured for inserting, using the processor, the residual signal into the data stream.
10. The encoder of claim 9 , wherein the data stream generator is further configured for inserting first position information indicative of the rendered portion of the first view into the data stream.
11. The encoder of claim 9 , wherein the data stream generator is further configured for inserting second position information indicative of the renderable portion into the data stream.
12. The encoder of claim 9 , wherein the determinator is configured for determining the renderable portion to generate the rendered portion based on depth/disparity map of a scene associated with the first and second views.
13. The encoder of claim 9 , further comprising a reconstructor configured for reconstructing, based on the prediction signal and the residual signal, a reconstructed reference signal to be used by the view predictor for block-based prediction.
14. The encoder of claim 9 , wherein the view predictor is configured for performing block-based disparity compensated prediction to obtain a reference signal for a renderable portion of a third view.
15. A method for encoding a multi-view signal representing a video into a data stream, comprising:
determining a renderable portion of a first view of the multi-view signal, by view synthesis based on a second view of the multi-view signal to generate a rendered portion of the first view;
performing a block-based prediction of the multi-view signal based on a previously encoded portion of the multi-view signal and the rendered portion of the first view, to obtain a prediction signal;
determining a residual signal between the prediction signal and the multi-view signal; and
inserting the residual signal into the data stream.
16. The method of claim 15 , further comprising inserting first position information indicative of the rendered portion of the first view into the data stream.
17. The method of claim 15 , further comprising inserting second position information indicative of the renderable portion into the data stream.
18. The method of claim 15 , wherein the renderable portion is determined based on depth/disparity map of a scene associated with the first and second views.
19. The method of claim 15 , wherein non-renderable portion of the first view is disjoint to the renderable portion of the first view.
20. A method for storing a data stream, comprising storing, on a digital storage medium, a data stream into which a multi-view signal representing a video is encoded by a method according to claim 15 .