Mentor
- 11,130
- 3,982
In the insights article, I did on how 8k video can be distributed at internet speeds we have usually access to these days, I mentioned LCEVC:
For more details see:
https://ieeexplore.ieee.org/document/9795094
It is a very effective technique with older codecs, but with newer codecs like VVC the gains are only small, with not using it often producing better results:
https://ieeexplore.ieee.org/iel7/76/4358651/10019298.pdf
As mentioned in previous posts, since LCEVC was standardised, AI has made significant gains. An AI varient of LCEVC will be shown to outperform LCEVC.
It uses an auto-encoder called CAESR:
https://arxiv.org/pdf/2202.00416
This is combined with an AI-infused varient of LCEVC:
https://www.tnt.uni-hannover.de/pap...ding_with_Spatial_and_Temporal_Prediction.pdf.
It outperforms LCEVC, except for a multilayer configuration.
Again, this post has introduced a number of new ideas the reader may not be familiar with. It will take a bit of time to assimilate, but already it is seen that AI outperforms single-layer LCEVC.
The next post will show how the scheme can be enhanced using the newer super-resolution techniques from a previous post and enhancements to EVC Baseline Inter coding.
For more details see:
https://ieeexplore.ieee.org/document/9795094
It is a very effective technique with older codecs, but with newer codecs like VVC the gains are only small, with not using it often producing better results:
https://ieeexplore.ieee.org/iel7/76/4358651/10019298.pdf
As mentioned in previous posts, since LCEVC was standardised, AI has made significant gains. An AI varient of LCEVC will be shown to outperform LCEVC.
It uses an auto-encoder called CAESR:
https://arxiv.org/pdf/2202.00416
This is combined with an AI-infused varient of LCEVC:
https://www.tnt.uni-hannover.de/pap...ding_with_Spatial_and_Temporal_Prediction.pdf.
It outperforms LCEVC, except for a multilayer configuration.
Again, this post has introduced a number of new ideas the reader may not be familiar with. It will take a bit of time to assimilate, but already it is seen that AI outperforms single-layer LCEVC.
The next post will show how the scheme can be enhanced using the newer super-resolution techniques from a previous post and enhancements to EVC Baseline Inter coding.