Hello Michael, Actually my codec is very lossy-oriented, and generally what works in lossy, doesn't really work in lossless... Lossy and lossless are two different things...
For now at first glance, I don't see really good ideas from lossy NHW Project to improve lossless FFV1... but I will study it deeper.Even my entropy coding schemes are for lossy scheme... What I can advice, but you certainly already know it, I have read recently (on the encode.ru forum) that some developers have implemented good new image models (context and predictors) to the paq8px compressor which gave them even better compression, but I think it is complex.The paq8px compressor has then very good results on lossless image compression but is slow... Maybe if you are interested in these new image models, you can contact them. If I go through new good ideas while developing the NHW Project that could be useful to FFV1, I will let you know. Many thanks again! Kind regards, Raphael Canut Le ven. 5 oct. 2018 à 01:24, Michael Niedermayer <[email protected]> a écrit : > Hi Raphael > > Thu, Oct 04, 2018 at 03:59:08PM +0200, Raphael Canut wrote: > > Hello Michael, > > > > Many thanks for your answer Sir! > > > > > It interests me a lot that the NHW Project algorithm would be studied for > > the great FFV1 codec! > > I think it would only be studied if someone (like for example you) would > do that study, that is compare it to current ffv1 and suggest it as an > algorithm for a future version. Of course you would only do that if/when > you see that as making sense. For example when it is better than the > current code, supports lossless and arbitrary image sizes ... > > What i wanted to say with my reply is more that suggestions to extend > ffv1 in ways that improve it would be welcome. > > > > > > Just first I wanted to point out that the NHW Project is very > experimental > > for now.For example there is no lossless mode and it is now only for > > 512x512 image size... > > > > > Do you plan to add to FFV1 a lossy intra-coded picture mode? > > That would only make sense if that is either a very simple & natural > extension > to existing code or if it adds significant value over other existing > codecs. > > > > > > Maybe first and before going deeper, you can try the NHW Project and see > if > > in its current state it can be interesting for FFV1? > > > > Many thanks again! > > Cheers, > > Raphael > > > > > > > > Le jeu. 4 oct. 2018 à 14:37, Michael Niedermayer <[email protected]> > a > > écrit : > > > > > Hi Raphael > > > > > > On Wed, Oct 03, 2018 at 06:42:01PM +0200, Raphael Canut wrote: > > > > Hello, > > > > > > > > Maybe it could be interesting for professional people of this list, I > > > have > > > > improved the NHW Project image compression codec. > > > > > > > > This new version is better and I have added very high compression, > which > > > is > > > > very competitive with x265 (HEVC).Maybe this version could be now > > > > interesting for the IETF video codec working group... > > > > > > > > For reminder, the NHW Project has more neatness than x265 (HEVC) and > I > > > find > > > > it visually more pleasant, it is royalty-free, and it is at least 50x > > > > faster to encode and at least 15x faster to decode than x265 > (optimized > > > > HEVC)! So it is meeting the criteria of the IETF video codec. > > > > > > > > You can find more at: http://nhwcodec.blogspot.com/ > > > > > > > > Sorry again if you find that I am "spamming" the list (do not > hesitate to > > > > let me know), but it is extremely difficult for me to reach the > Industry > > > > for my codec. > > > > > > I cant speak about NHW (as i have not looked at it yet) or for IETF > > > but somewhat on topic and related, if you are interrested in > > > working on a IETF "standarized" video codec". As the author of the > > > original FFV1 spec. I and i belive "we" at IETF cellar would be very > > > interrested > > > in seeing work done to improve future iterations of FFV1. And while > > > FFV1 is not a wavelet codec really i think there are no real > constraints > > > of what a future v5, v6 or whatever might do, it could use a completely > > > different algorithm if the people on IETF-cellar feel thats the right > > > way forward. > > > I had always imagined long before cellar-IETF existed that future > > > revissions > > > of FFV1 would be designed based on some sort of competition between > > > different algorithms being tested with the aim being simplicity, > > > high compression, speed and losslessness. > > > Also image coding and intra only coded video are very closely related, > > > its basically almost the same problem. > > > > > > Thanks > > > > > > [...] > > > > > > -- > > > Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB > > > > > > No human being will ever know the Truth, for even if they happen to > say it > > > by chance, they would not even known they had done so. -- Xenophanes > > > > > > _______________________________________________ > > video-codec mailing list > > [email protected] > > https://www.ietf.org/mailman/listinfo/video-codec > > > -- > Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB > > No human being will ever know the Truth, for even if they happen to say it > by chance, they would not even known they had done so. -- Xenophanes >
_______________________________________________ video-codec mailing list [email protected] https://www.ietf.org/mailman/listinfo/video-codec
