This Vector-Based Video Codec Promises Tiny, Resolution Independent Movies

Apple Maps still needs work if it wants to supplant Google's offering, but one thing it has gotten right is the use of vector data over raster images. When it comes to downloading new information or zooming in and out, Apple Maps is the superior product. So why not use vectors for say, encoding video, where its space-saving and quality-preserving benefits would be just wonderful? Don't worry, science is onto it.

A press release over on the University of Bath's website states that "researchers" at the institution believe their vector-based video codec could bring about the "death of the pixel within the next five years". While, I'm sure the software engineers behind H.264 will have something to say about that, storing video information as points (and some colour data) rather than frames of pixels sounds like a better approach. Admittedly, psycho-visual optimisations and predictive algorithms have made the codecs we currently have very efficient, but no one is going to say no if an alternative approach yields fruit.

Before you ask, yes, there's a 30-second video to demonstrate the progress the team has made so far. It features various clips of raw footage followed by encoded footage, either superimposed or side-by-side. As you'll see, the codec transforms the pixel information into "coloured contours", which the release states will serve as the basis for refining the codec. It goes on to mention that the team is trying to drum up industry support and then, well, that's about it, really.

I think it's unlikely we'll be laughing at the crudeness of XviD and H.264 by 2017 and rubbing ourselves down with vector codec deliciousness, but I am completely in the love with the idea of downloading crazy tiny videos that scale to every possible resolution and are completely free of compression and motion artefacts.

[University of Bath, via DVICE]



    Sounds like an awesome step forward, if they can pull it off. How tiny are we talking? A few MB?

    Doesn't google maps also use vector graphics on iOS app now?

    Apple Maps still needs work if it wants to supplant Google’s offering, but one thing it has gotten right is the use of vector data over raster images. When it comes to downloading new information or zooming in and out, Apple Maps is the superior product

    Oh my god. That has to be the Most stupid sentence I have ever seen.
    Yes, google maps on iOS used bitmaps; but google maps on browsers and on android/windows 8 phones has used vector mapping since 2009!!

    If you don't update your article, you will lose all of your credibility.

      So they're not actually incorrect. You just want to bring Android into an article that has nothing to do with phones. They're not going to lose credibility because no one cares about phone wars anymore.

      He's an Apple fanboy, cut him some slack. He's not expected to know much about other tech.

        Uh, no. I made the comparison because it is one many people will get immediately.

        Can we stop with the "so and so must be a fanboy" stuff? It's really quite tiring and pointless.

      Not to mention the new Google Maps app for iOS now also utilises vector maps. It only use bitmaps for satellite view.

      I don't see why it needs to be updated. The comparison, made in the context of mobile, is valid.

        Not if what the other commenters have said is true.
        Looks like Gizmodo journalism is as bad as their sister sites'.

    Perhaps their claims would at least appear realistic if their only sample video wasn't pixel-based. Where is some vector video player and a sample vector video.

    Their claim kind of reminds me of the claims that fractal compression made last century where fractal compression gives you images that become resolution independent.

    If it wasn't so sad, this would be funny.
    This exact concept was conceived and documented by me (as Y-Planar Compression) in 2005, but my then business partner was too thick headed to understand that he was the business, and I was the IP.
    Bad luck to me I guess.

    There are always trade-offs though. I doubt it'd be scalable to any size because where would the compression be then? Vector based images involve a lot of simplification. You'd have to have that simplification at some level, even if it was time based rather than image based, as in interpolation between changing areas or something like that...?

    Last edited 15/12/12 5:21 pm

      @ozoneocean : Efficiency - based on the original documents - I was aiming for 150:1 to get broadcast distribution quality - without the artefacts of traditional raster-based codecs. You can get better, but the trade-offs in temporal or spatial compression are hard to hide!

    Vector codecs could be awesome at encoding cartons and anime, but they'll certainly be crap at preserving detail on anything with texture. Not expecting any h.264 challengers here, but if like to see a wider range of example video, at a decent resolution.

    And yeah, your Apple Maps comments aren't making iOS users look any less stereotyped.

    Well, the things to note here are:
    1. I'm not sure the video files are going to be small
    2. It will most likely require a massive increase in computing power
    3. There are a lot of other benefits, like the ability to render it differently depending on processing power etc.. a pixel isn't a pixel any more, it's dependant on the algorithm to decode it. eg: you may simply use half the vectors, or all of the vectors with different point counts (think lower res 3d models in games)
    4. streaming video based on this would open up a whole other range of benefits, for example, dynamically scaling up/down the resolution / vector count as discussed in point 3
    5. It is an awesome idea.

    Someone already mentioned the concepts of "Lytro camera" and "3D video" right?
    Ok, good. Because the concept of recording video that can be moved around in 3D, and is scalar, is pretty awesome.

Join the discussion!

Trending Stories Right Now