Hi, Keep up the good work! The file sizes of Theora's output is amazing! I have just joined the list, and would like to contribute to the Theora/Vorbis project on the mathematics side, but I don't know what is required. As of now I have the basics in mathematics of the Danish Technical University. So I was wondering if anyone can tell me, what sort of math that is the central of Theora and Vorbis, so I can choose courses based on that? My guess it that DCT is the core...? Martin --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.
On 6 Dec 2003 erivy7302@sneakemail.com wrote:> Hi, > > Keep up the good work! The file sizes of Theora's output is amazing! > > I have just joined the list, and would like to contribute to the > Theora/Vorbis project on the mathematics side, but I don't know what is > required. > > As of now I have the basics in mathematics of the Danish > Technical University. > > So I was wondering if anyone can tell me, what sort of > math that is the central of Theora and Vorbis, so I can choose courses > based on that? > > My guess it that DCT is the core...?Check out my VP3/Theora document at: http://home.pcisys.net/~melanson/codecs/vp3-format.txt The section "Underlying Coding Concepts" gives a brief overview of each mathematical concept used in Theora coding. The DCT is surely the most complicated one presented. However, there is also more math involved in encoding when you cover topics such as data rate control and psycho-visual models. For Vorbis, take a gander at the official spec: http://www.xiph.org/ogg/vorbis/docs.html For some reason, the coding concepts of audio coding tend to be more complicated than those for video coding. -- -Mike Melanson --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.
On Sat, Dec 06, 2003 at 05:24:16PM -0000, erivy7302@sneakemail.com wrote:> Keep up the good work! The file sizes of Theora's output is amazing!Glad you like it.> I have just joined the list, and would like to contribute to the > Theora/Vorbis project on the mathematics side, but I don't know what is > required.Timothy (derf) would be a better one to answer this; my training is in physics. It's all terribly applied and half-engineering of course. Yes, the MDCT is a big part of both Theora and Vorbis. Information theory is the basis of all compression technology so you should be familiar with that as well. Linear algebra is a useful way of thinking about what can be done and how to do it. Pretty much all compression technology amounts to 'transform the data to concentrate non-randomness, then efficiently entropy-code that'. For lossy codecs like Theora and Vorbis, the transformation step isn't completely reversible in a clever way so that you don't store information that's perceptually difficult to notice. Other helpful things are anything to do with signal processing and spectral theory, Green's functions. Wavelets too, although they're not currently used in the vorbis or theora designs. And general applied scientific and computing maths courses can be good. (Depending on the curriculum; they can also be really dumb.) Hope that's of some help, -r --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.
> Check out my VP3/Theora document at: > http://home.pcisys.net/~melanson/codecs/vp3-format.txt > The section "Underlying Coding Concepts" gives a brief overview of each > mathematical concept used in Theora coding. The DCT is surely the most > complicated one presented.That sure it well written and documented! I haven't read it all, will do it the following days. I find it very interesting as you give examples along the way!> However, there is also more math involved in encoding when you > cover topics such as data rate control and psycho-visual models.Have this been documented yet?> For Vorbis, take a gander at the official spec: > http://www.xiph.org/ogg/vorbis/docs.htmlThanks.> For some reason, the coding concepts of audio coding tend to be more > complicated than those for video coding.I wouldn't have guessed that=) Martin --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.
> Yes, the MDCT is a big part of both Theora and Vorbis. Information theory is the basis of all > compression technology so you should be familiar with that as well. Linear algebra is a useful way > of thinking about what can be done and how to do it. Pretty much all compression technology > amounts to 'transform the data to concentrate non-randomness, then efficiently entropy-code that'.Do you know where I can read about entropy-coding? I have searched Google and found this presentation, which is very illustrative with examples: http://bmrc.berkeley.edu/courseware/cs294-3/fall95/lectures/94avcomp2.ps But I can hardly learn from that.> Other helpful things are anything to do with signal processing and spectral theory, Green's > functions. Wavelets too, although they're not currently used in the vorbis or theora designs. And > general applied scientific and computing maths courses can be good. (Depending on the curriculum; > they can also be really dumb.)Wavelets you say? I have choosen a course on wavelets next semester. It have rumours of being tough. The book that I have bought for the course is "A First Course on Wavelets" ISBN 0-8493-8274-2. Do you think I will be able to use this knowledge to code compression algorithms? Using Green's funtions and wavelets; will that give a better preformance/size than Theora have now? And if so, is it considered to be used later on? Thanks. Martin <p><p><p><p><p><p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.
> MDCT is Vorbis only. Theora is based on a 2D 8-point Type II DCT (e.g., > a regular DCT). For the large majority of the work, you don't really > even need to understand how this works. The transform in use is fixed, > and our VP3 compatibility goal will not let us change it. You can just > treat it as a black box.What about in the future? Must later versions of Theora always be backward compatible with VP3? Or will Theora be renamed when it is not compatible with VP3 anymore doe to competition from other codecs?> If I were picking courses, I'd want some advanced linear algebra (e.g., > something that goes beyond the introductory course method of just > talking about matrices, eigenvalues, and eigenvectors, etc., and > develops the abstract ideas of what a vector space is, an inner product > space, etc.), and at least some simple singal processing (know what a > Fourier transform is, its connection with the DCT, understand at least > the basics of the frequency domain). If your university has a > compression/information theory course, it might certainly be > interesting, but Theora doesn't use much beyond RLE and Huffman codes.An advanced linear algebra course sounds interesting indeed! Can you give an example where linear algebra is heavily used in Theora?> After that there is a ton of specialized domain literature that needs to > be read and understood. I doubt many universities have actual classes on > these things, and if they do, they probably don't delve very deep. Don't > let this scare you; most of this part can be done "on the job" as it were.Well, I already now have some clues; signal processing and spectral theory, advanved linear algebra, Information and compression theory. Do you know titles of this specialized domain that you speak of? Thanks. Martin <p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.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.