Hi,
I have been on a documentation kick recently. I finally updated my
VP3 format document tonight:
pcisys.net/~melanson/codecs/vp3-format.txt
v0.2: October 9, 2003
- expanded section "Reversing the DC Prediction"
- added Appendix B: Theora Differences
Feel free to look it over and point out any glaring mistakes. I can't wait
to see how much sense it makes to me in the morning. But I completely
revised the DC prediction section so that it is almost entirely separated
from any specific implementation details (earlier weakness).
Thanks...
--
-Mike Melanson
--- >8 ----
List archives: xiph.org/archives
Ogg project homepage: xiph.org/ogg
To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body. No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.
> Feel free to look it over and point out any glaring mistakes. I can't wait > to see how much sense it makes to me in the morning. But I completely > revised the DC prediction section so that it is almost entirely separated > from any specific implementation details (earlier weakness).Didn't read too deeply yet, but a few comments on the Theora differences section: 1) We plan to fix the frame orientation to conform to that used with VP3 before the beta (most likely in an alpha3). We want to preserve the ability to transcode from VP3 losslessly, and there's no real reason to keep two different orientations around. 2) Theora stores the prefix-free codes directly, in the form of a full binary tree (this is a change from alpha 1, where the histograms used to generate the prefix-free codes were stored). 4) Theora removes some of the unused bits in the VP3 frame header. There are still 3 0 bits in keyframe headers (keyframe type, which is always 0, and 2 unused bits), but the others are gone. --- >8 ---- List archives: xiph.org/archives Ogg project homepage: xiph.org/ogg To unsubscribe from this list, send a message to 'theora-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Happy Leap Day,
I finally wrote some more of my VP3 document:
home.pcisys.net/~melanson/codecs/vp3-format.txt
Changelog:
v0.3: February 29, 2004
- expanded section "Unpacking The Macroblock Coding Mode Information"
- expanded section "Unpacking The Macroblock Motion Vectors"
- added Appendix: Macroblock Coding Mode Alphabets
It may not seem like much, but gosh, this stuff takes a long time to
write.
Timothy T.: I haven't lost all of your useful updates. But I am
unfortunately too tired to include them on this update. They will make it
in eventually.
Thanks...
--
-Mike Melanson
--- >8 ----
List archives: xiph.org/archives
Ogg project homepage: xiph.org/ogg
To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body. No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.
Hi Mike, this is a great spec. yes these things take a lot of time to write! would you be interested in formulating it as an internet draft towards an rfc? I'd be happy to do the required reformatting and boilerplate work. You wouldn't need to do any extra work other than agree :) I'm currently at the ietf meeting in seoul (enjoying muchos wireless :) so I'm well keen for it. Is anyone else on this list there? Conrad. On Sun, Feb 29, 2004 at 10:37:40PM -0700, Mike Melanson wrote:> Happy Leap Day, > I finally wrote some more of my VP3 document: > home.pcisys.net/~melanson/codecs/vp3-format.txt > > Changelog: > v0.3: February 29, 2004 > - expanded section "Unpacking The Macroblock Coding Mode Information" > - expanded section "Unpacking The Macroblock Motion Vectors" > - added Appendix: Macroblock Coding Mode Alphabets > > It may not seem like much, but gosh, this stuff takes a long time to > write. > > Timothy T.: I haven't lost all of your useful updates. But I am > unfortunately too tired to include them on this update. They will make it > in eventually. > > Thanks... > -- > -Mike Melanson > --- >8 ---- > List archives: xiph.org/archives > Ogg project homepage: xiph.org/ogg > To unsubscribe from this list, send a message to 'theora-dev-request@xiph.org' > containing only the word 'unsubscribe' in the body. No subject is needed. > Unsubscribe messages sent to the list will be ignored/filtered.--- >8 ---- List archives: xiph.org/archives Ogg project homepage: xiph.org/ogg To unsubscribe from this list, send a message to 'theora-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
> Timothy T.: I haven't lost all of your useful updates. But I am > unfortunately too tired to include them on this update. They will make it > in eventually.Especially the one about Theora no longer having the opposite orientation that VP3 has (at least in the CVS version). The other comment I'd make is that the DCT tokens are coded in terms of Huffman codes. The fact that the VP3 code re-derives the Huffman codes corresponding to a set of histograms every time you run it is mostly irrelevant. The histograms are fixed (in VP3), and so the codes are fixed. In Theora, the codes are specified in the bitstream directly, not in terms of a histogram. Also, I don't remember if anyone ever asked before, do we have permission to incorporate pieces of this document (with attribution) into our own Theora bitstream spec.? --- >8 ---- List archives: xiph.org/archives Ogg project homepage: xiph.org/ogg To unsubscribe from this list, send a message to 'theora-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.