![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
xiphmont
Another new demo, another new technique specific to Daala: frequency domain prediction of the chroma planes from the luma plane!
Predicting the chroma planes from the luma plane isn't a brand-new idea. Still, we're both the only codec to actually be deploying it, and we're doing it entirely in the frequency domain (which is novel).
Replacement for JPEG?
Date: 2013-10-20 08:34 am (UTC)I just wonder whether you can create a new lossy image format (as a "sister" format of Daala). It seems to me that it would be much more efficient than a JPG if all relevant techniques that will be used in Daala would be taken...
What do you think?
Petr
Re: Replacement for JPEG?
Date: 2013-10-25 02:29 am (UTC)I don't think JPEG is especially screaming out for improvement. Yes, you can do better than JPEG and perhaps a Daala intra frame would. But it would have to be quite a bit better for anyone to want to shoulder the substantial cost of replacing JPEG.
Second, Google is already doing something similar with WebP, which started out as a WebM (VP8) intra frame. I'm not so interested in how much more efficient it is (although Google might be), it's the additional alpha channel that makes WebP compelling to me.
Re: Replacement for JPEG?
Date: 2013-10-25 04:21 pm (UTC)Another great and impressive demo!
Concerning JPEG, I have compared it with WebP (VP8 intra frame) and x264 intra.Yes, they are a little better than JPEG (JPEG is still visually good), but they are also slower.WebP seems compelling because of the alpha channel, (web-centric) lossless compression, animation...
Just a little off-topic, what about wavelet codecs like for example the nhw codec I have posted on the Xiph Theora channel? The nhw codec would have similar quality as JPEG (sometimes better, with advantages and drawbacks) and would be x2/3 times faster.Could this be an interesting feature (speed)?
Just a last question, for the Daala frequency prediction, did you study the SPIHT model used in most wavelet codecs?
RC
no subject
Date: 2013-10-21 05:20 pm (UTC)This is strange. Was it measured with same encoded size, or is it apples to persimmons comparison again, like in demo 1?
How does CfL interact with spatial chroma prediction?
no subject
Date: 2013-10-25 02:32 am (UTC)This is measured in the full encoder (including other prediction and coding) accounting for size.
>How does CfL interact with spatial chroma prediction?
Did you mean intra-style chroma prediction (we can't do any prediction in the spatial domain in Daala)? It replaces it entirely.
Or did you mean spatial CfL and spatial chroma prediction in h264/HEVC? (also a complete replacement.)
no subject
Date: 2013-10-25 06:26 pm (UTC)yes.
I take it you tried to transmit the B coefficient explicitly, rather than inferring it?
no subject
Date: 2013-10-21 07:19 pm (UTC)no subject
Date: 2013-10-25 02:34 am (UTC)Interesting comparison
Date: 2013-10-29 05:21 pm (UTC)http://maxsharabayko.blogspot.ru/2013/10/next-generation-video-codecs-hevc-vp9.html
Re: Interesting comparison
Date: 2014-02-07 10:27 pm (UTC)Daala right now is not fit for any form of comparison. Although the encoder and decoder work, they do not produce anywhere near competitve output compared to completed codecs.
Chroma subsampling
Date: 2013-12-31 01:25 pm (UTC)Regarding chroma subsampling, will Daala support 4:2:2 and 4:4:4 as well? It's not really useful for video, but as first anon, I'm looking for candidates for a JPEG replacement and webp, based on VP8, is lacking in that regard.
Have a great 2014.
Re: Chroma subsampling
Date: 2014-02-07 10:29 pm (UTC)