A pipe dream came true? A new vector-based codec has been developed

[birdie]

Is the pixel about to die?

11 December 2012

Researchers launching a new vector-based video codec are claiming their work will lead to the death of the pixel within the next five years.

The team behind the project, consisting of the University, Root6 Technology, Smoke & Mirrors and Ovation Data Services – are now looking for industry buy-in to the research to expand its potential applications. The codec was launched at the CVMP 9th European Conference on Visual Media Production held at Vue Cinema in Leicester Square, London.
A new vector-based video codec could signal the end of the pixel.

A new vector-based video codec developed at Bath may signal the death of the pixel.

Digital pictures are built from a rectangular grid of coloured cells, or pixels. The smaller and closer the pixels are together, the better the quality of the image. So pixel-based movies need huge amounts of data and have to be compressed, losing visual quality. They are also difficult to process.

The alternative, a vector-based format, presents the image using contoured colours. Until now there has not been a way to fill in between the colours at the quality needed for professional use. The Bath team has finally solved this problem.

A codec is a computer programme capable of encoding or decoding a digital video stream. The researchers at Bath have developed a new, highly sophisticated codec, which is able to fill between the contours, overcoming the problems previously preventing their widespread use. The result is a resolution-independent form of movie and image, capable of the highest visual quality but without a pixel in sight.

Professor Phil Willis, from the University’s Department of Computer Science (http://www.bath.ac.uk/comp-sci), said: “This is a significant breakthrough which will revolutionise the way visual media is produced.

“However, to accelerate this project we’ll need companies from around the world to get involved. At the moment we’re focusing on applications in post-production and we’re working directly with leading companies in this area, however there are clear applications in web, tablets and mobile which we haven’t explored in detail yet.

“Involvement from a greater variety of companies with different interests will extend the project in a variety of ways and increase the potential applications of this game-changing research.”

To learn more about the project and to see the video demonstration given at CVMP visit: http://www.cs.bath.ac.uk/vsv

Something tells me this new codec will work well only with anime type movies.

[Pomegranate]

It would seem well suited to anime and animation movies indeed. But how exactly does one shoot a live action movie in a vector format?

[TheElix]

I came here to post these news but somebody already did. If it really works it's a revolution.

[schweinsz]

Is there some .pdf papers to introduce it's algorithm? I assume that there is some common thing between their method and the sparse sampling.

[iwod]

It would seem well suited to anime and animation movies indeed. But how exactly does one shoot a live action movie in a vector format?

I never believed in the so call Vector Based Video. But i never thought about Anime!

Could i Theoretically mix Vector and Pixel Based together for best of both world?

[mandarinka]

Contrary to popular belief, anime usually isn't vector-like. Sure there are probably cases that resemble The Simpsons or Futurama (simple outlines with completely uniform solid color fill). But usually there are painted/drawn backgrounds and even solid color areas use some texturing/noise to avoid looking bland. In other words, those vectorising approaches can't do anything breaktrhough here, without the output looking terrible. With superflat 2D computer animation, it's probably inter (well intra too) prediction that is the silver bullet, anyway :)

[mastrboy]

How well does it deal with grain preservation on anime sources? :P

[ckmox]

Something tells me this new codec will work well only with anime type movies.

did you check the sample video work they link to that article? here - http://www.cs.bath.ac.uk/vsv/contour.mov

looks epic, and i thought Fractal Video Compression is the best of both worlds (Raster+Vector) but looks like Vector alone can do Raster/Pixel/Bitmap strength too now, fingers cross on when will this be release, i hope it will become open source, or google should support the research and development of this vector-base codec more

[schweinsz]

did you check the sample video work they link to that article? here - http://www.cs.bath.ac.uk/vsv/contour.mov

looks epic, and i thought Fractal Video Compression is the best of both worlds (Raster+Vector) but looks like Vector alone can do Raster/Pixel/Bitmap strength too now, fingers cross on when will this be release, i hope it will become open source, or google should support the research and development of this vector-base codec more

I don't take this. The fractal based method is not that powerful. Fractal is to use the non local correction in its nature.
Incorporating the non local correction can't give much gain.
You can read some template match based or block match based intra coding papers. Alrough it could give modorate gain for the H.264 intra coding(say about 10%), That could only give very limited gain on the HEVC.

[vivan]

did you check the sample video work they link to that article? here - http://www.cs.bath.ac.uk/vsv/contour.mov

looks epicLooks like awfully encoded video (it's 2012 now, lol, not 2002). And it doesn't really show anything.

And picture on their page looks very nice too. Like 256-color gif image.

[LoRd_MuldeR]

Looks like awfully encoded video (it's 2012 now, lol, not 2002). And it doesn't really show anything.

That clip on their web-site is Baseline H.264 video, probably encoded with the (crappy) QuickTime encoder.

Therefore doesn't say anything about their compression scheme!

We'd either need a bitstream plus a decoder. Or at least a lossless version of the output from their decoder.

And picture on their page looks very nice too. Like 256-color gif image.

Believe it or not. It's because it is a 256-color GIF image :eek:


Unless they can provide an encode of one of the "standard" test clips (e.g. Park Joy (http://media.xiph.org/video/derf/y4m/park_joy_1080p50.y4m)), which can compete with state-of-the-art H.264 encoders at a reasonable bitrate, I wouldn't put too much hope into that ;)

[mandarinka]

Well in any case, them producing a *usable* codec with great quality and (truly competetive) compression performance is way less likely than it being hype and the actual method sucking. It's not an easy problem, and there is Murphy's law too.
In short, it's naïve to expect anything from this.

[ckmox]

so much negative expectations on this thread but im hopeful (and maybe im hopeful since im not expert like the people here) but here is more info about this with PDF of its technical things too that you can find while reading this entirely - http://www.extremetech.com/extreme/143130-vector-vengeance-british-researchers-claim-they-can-kill-the-pixel-within-five-years

[jethro]

"Within five years" everything will be in 5 years apparently, it just needs more funding

[CruNcher]

It looks like a massive visual improvement (Look & Feel) if the examples are in the end @ least the same size :)
Though 5 years i guess it will be interesting HEVC is close to the end of being enhanceable ;)

[birdie]

so much negative expectations on this thread but im hopeful (and maybe im hopeful since im not expert like the people here) but here is more info about this with PDF of its technical things too that you can find while reading this entirely - http://www.extremetech.com/extreme/143130-vector-vengeance-british-researchers-claim-they-can-kill-the-pixel-within-five-years

You definitely want to read the comments to your link. They are quite insightful. I especially liked the references to NASA, aliens, quantum computers, 200 (sic!) cores laptops' CPUs, and 600GB of RAM (I guess 640GB of RAM will be enough for everybody).