Two hour movie, 25 frames per second. All in 16MB!?!

[Ed Hansberry]

<a href="http://hollywood.org/DGS/index2.html">http://hollywood.org/DGS/index2.html</a><br /><br />Guillaume Defossé has developed a compression algorithm he claims will allow you to view a two hour movie from a 16MB storage card on your cell phone or PDA, though I suspect a PDA movie would be larger to allow for the bigger screen. Still, even if it ballooned to 32MB for a Pocket PC, who cares? This would be awesome. Just think, Marlof would be able to fit roughly 16 movies on that <a href="http://www.pocketpcthoughts.com/forums/viewtopic.php?t=1463">512MB SD card</a> you just know he is itching to order.<br /><br />Thanks to Michael van Oosten for the link.

[denivan]

This guy is from Belgium. I read about his system two years ago in a magazine, but he never got much attention. He allways was pretty misterious about his invention so most of the people who knew about this were sceptical. I'm very interested to see what's the quality...

[JMountford]

No offence guys but as somewhat pointed out by previous poster, this is old news. Now if adopters of this encryption algorithom start popping up at a rapid pace that would be news.

[Ed Hansberry]

No offence guys but as somewhat pointed out by previous poster, this is old news. Now if adopters of this encryption algorithom start popping up at a rapid pace that would be news.

How old could it be? He is running it on a Nokia 9210 and did limited demo's March 18 at CeBIT. He may have been working on it for years before that, but I think the "news" is he may have a viable product - he is estimating 4-5 months to market for the 9210.

Lemme know if you want me to filter out stuff that isn't up to the minute. I can just assume everyone else has already seen it. :roll:

[denivan]

Well, he has had 'this invention' for a long while now, but for some reason he never cashed in, which made everybody in our country even more suspicious. The company he now works with is also located in Belgium, so I seriously doubt that his invention is real. If it trully was real, he would have sold it for billions to nokia or something, instead of working with some crappy unknown company that has difficulty convincing people of its product. By the way, if you download that stupid videoclip, then it all says enough...that clip hardly looked like 25 fps to me.

In my opinion this guy is just a fake.

[njb42]

The article mentions that 11MB of the 16MB card is used by the OS and applications. So it appears he's claiming to have compressed a 2-hour movie into 5MB?

I dunno. I've done some coding of experimental video compression algorithms. If he really has accomplished what he claims, my hat is off to him.

The other question is, how long does it take to compress the movie? Generally the smaller the output file, the more CPU time it takes to crunch through the compression. I wouldn't be surprised if his algorithm took several hours even on a very fast machine.

[bchristian]

If it sounds too good to be true... It probably is. That is what I think anyway. It seems like we get these kind of claims every few months about compression. I don't see how we can get much better compression than we have now.

If he really had a breakthrough then we would have seen something using it by now. And even if he has it compressing better than current algorithms I bet it is like njb42 says and it takes hours or maybe even days to compress a single movie on a Pentium 2.53Ghz.

It reminds me of the "miracle" carburaetors of days gone by. I used to hear about these every once in a while with claims like "100 miles per gallon or more!". I guess in our age the claim is "Over 2 hours of video and audio in 5MB!"...

Yeah, right. I will believe it when I see it, and can examine it close-up.

[CR]

A 108 min. movie at 25fps in 4MB... that's like 1 frame in every byte plus the audio. That's never going to happen if you want any viewable quality in your video. Looks to me like he's just streaming it over a network. On the board in the background, it says 75kb/s ( unless that's 15 ) which comes out to roughly 60 MB. At 15 Kb/s it comes out to 12 MB, but there is no point in trying to watch that espescially when you tack on the audio stream. I guess it would be okay if the resolution was 20x20 ;)

[Ed Hansberry]

I too am beginning to get suspicious. Check out his FAQ at http://hollywood.org/DGS/FAQ.html - look at this question:
Question: Is DGS loss-less compression ?
Answer: We are not completely sure for the moment but we think it is. The reason is that we can compress (better : reduce) (1) motion/picture/audio, (2) Data, but also (3) software (binary executable code). An example: we reduce a well-known software packet of about 450 MB to 20 MB (stored on a CD-business card). When installing it runs on the 20 MB version (thus not decompressing the 20 MB back to 450 MB).

Huh? "We think it is?" What does that mean? Either you can decompress it to its full glory or you can't. Not that hard to test. :roll:

[denivan]

For your information i looked up the company 'Millenium Gate' in 'de gouden gids' (that's the belgian yellow pages) and found nothing. I also looked it up on vanhecke.com (a company listing site, our home companies are listed here, one of them was just founded last year, so vanhecke.com is up to date). So on both sites I can find nothing about this Millenium Digital NV , at the moment I have exams, but I would like to go to the national bank depot, to see if this company really exists.

Another thing on this page is funny :
http://hollywood.org/DGS/warning.html

"Please keep in mind that special hardware adaptions are needed for DGS-applications. "


What does that mean? That a 1Gb microdrive needs to be built in into a Nokia 9210 ? ;-))

I'm almost embarassed to be belgian ;) Btw, in earlier interviews he stated that he is a home made inventor and is not an engineer of some sort...

[denivan]

So on both sites I can find nothing about this Millenium Digital NV

Of course that should have been Millenium Gate ;-)
Anyway I've also done a search on vanheck.com looking for just the word 'millenium', or defossé or guillaume and nothing turned up.

[William]

Of course that should have been Millenium Gate

No, it should have been Millennium Gate... :)

[denivan]

Damn, it must be typo night tonight ;-)

Anywayz, i've been browsing around hollywood.org and so far it seems that Dirk Laureyssens, Chief Operating Officer of Millennium Gate is the same guy who invented this : http://www.cricro.com/

Anywayz, I'm really interested to see how this thing will work out ;-)

It looks strange, but by what we can read in their FAQ it seems like they don't fully understand their own technology...

[mvoosten]

On the FAQ thing: You are probably refering to this line:
"Question: Is DGS loss-less compression ?
Answer: We are not completely sure for the moment but we think it is."

Before posting this 'news' item I really thought hard if the so called invention is real or a scam before posting.
The reason why I did post is is because it's very plausible that it is no scam.

The lines "we are not completely sure" I read as follow:
If you read carefully you see that they refer to their 'compression' technology as a new 'language'. So drawing the line it possibly is some form of describing language that describes the content. Using various techniques that will describe objects and interpret them correct may indeed lead to a massive compression of data.
Dor example: If you are reading a book and in a line you read "On the wooden round table in the small dark room there was a big red ugly lamp from the 60's" You immidiately have a picture of the scene in front of your eyes without telling you in detail bit by bit how you should render the scene in front of you. That is, you know roughly how the table, lamp and room will look and that the lamp doesn't emit much light or is most likely out.
That is because you interpret the words in a way that makes sence to you.

I guess something like this is the basis of this invention and if you think of it could make sence.

Now on the question why they are so vague on the losseless compression or not: They are probably not sure if the description of the 'content' is providing enough information to get the whole picture or that it leaves blanks to interpret in your own way.

What about the various format's you would probably say? Well, my guess is that it uses an OS extention (hence the site mentions it between the lines) that interprets the DGS package and describes it back to the player/application that is used how the content should look.

So.. now it's I guess time for the PocketPC community to speak as I saw they where looking for a possible launch platform and looking at their comment it seems they are having some issues with Nokia:
"It is however not sure that Nokia will be our preferred partner. It is possible that we will work with another mobile phone manufacturer or even start with a PDA."
Wouldn't it be great if the PocketPC platform was the first.
I would say: Mail them!!! ([email protected])

Well that's my thoughts for now: He it's pocketpc THOUGHTS isn't it ;)

M.

[TomB]

Ed, thanks for the link to this guy. I like to keep an open mind, but the demonstration mentioned would have to have a datarate of about 3% of what is currently acceptable. As far as the MOV file, it IS 25fps and the "motion" captured does show change in individual frames, but other than the irregular searchlight sweeps, there is zero "movement" just a series of stills with no animation - just setup changes. The quality of the demo is terrible so there is no way to tell what this guy has.

I would say this is an ongoing and elaborate hoax, although I can't figure out why anyone would waste time on this. The website was registered almost five years ago by Dirk Laureyssens who appears on the photo page where it mentions that Defossé Guillaume could not attend ";-)"

What a major dissapointment and waste of time!

[denivan]

Quote : Dor example: If you are reading a book and in a line you read "On the wooden round table in the small dark room there was a big red ugly lamp from the 60's" You immidiately have a picture of the scene in front of your eyes without telling you in detail bit by bit how you should render the scene in front of you. That is, you know roughly how the table, lamp and room will look and that the lamp doesn't emit much light or is most likely out.
That is because you interpret the words in a way that makes sence to you.

I guess something like this is the basis of this invention and if you think of it could make sence.
/End quote

You say this could make sense? What you're actually saying here is like you feed a program some script that says a lamp should be displayed but not exactly how. In my opinion this is totally insane and i really don't understand how it could make any sense. The only thing that comes close to this concept is vector drawings (simple explained : the command would be : draw me a circle with a certain radius, instead of saying put a dot here, another there, another there, and so on...), and they hardly have the quality for full motion video.

Btw, the eure.net, hollywood.org domains are owned by the same man. Eventually you will allways come out at that dude who invented the game puzzle I posted above. I suggest you guyz one thing : think twice before you believe something, if you don't believe me, read this story about how the biggest companies in the us were scammed :

http://www.jacksonville.com/tu-online/stories/050502/met_9322821.html

Another thing, in earlier interviews he stated that a full size movie was about 30 Mb...so let's do some math :

a 108 minute movie at 25 fps would contain 162000 frames in total.
30 Mb is 31457280 bytes. If you divide this by the number of frames, then you would get about 194 bytes per frame. Please explain me how you can fit one frame (filling a total screen) of video in 194 bytes of data , that is insane. And then i didn't even count the data needed for the audio.

Untill I see some proof (and believe me, belgian magazines have been asking for some proof for long) i consider this a hoax.


Btw, In earlier statements he called it a new language because he converted the media (foto's, movies) into another language consisting of 27 characters (the number might be wrong). I've been looking for the magazine I had, but can't find it anywhere. Anyway, that was all he ever said about the 'technical part'.

Greetz (btw, read that full story i linked to about the hoax)

[PlayAgain?]

This was covered some time ago on www.my-communicator.com and, I think it is highly unlikely.

I'd love to be proved wrong...... but I doubt I will on this one.

[paulv]

a 108 minute movie at 25 fps would contain 162000 frames in total. 30 Mb is 31457280 bytes. If you divide this by the number of frames, then you would get about 194 bytes per frame. Please explain me how you can fit one frame (filling a total screen) of video in 194 bytes of data , that is insane. And then i didn't even count the data needed for the audio.


What you do need to keep in mind of course is that you don't compress each frame separately with video compression systems, you look for pixel changes. So (I'm most certainly not an expert on this) if you had only a few pixel changes in every frame you could do this fairly easy - shame about the audio though as speech has lot's of changes. You'd also have to describe the first frame in some detail to give a reference for further changes.

All-in-all I reckon this is either a hoax or a misinterpretation. My thought is that we are talking streaming video in which the amount of memory isn't really all that relevant apart from buffering and frame recreation.

I certainly got a laugh out of their suggestion that they could compress a 450MB program to 20MB and still run it. That's pure BS. You'd still need to de-compress it at some stage to get the source code which would slow processing to a crawl if you did it on the fly.

This is really as bad as any e-mail hoax. But still good for a laugh :-)

[Rob Alexander]

Guys, Denivan pointed you right to the answer of how this works. It's obviously based on Cricro technology!
http://www.cricro.com/images/Greencricro.gif
See, they just take the pieces and put them back together smaller. And since they're made of foam, you can just squash them to make it really small. Just imagine that this is video and that's how it works! Sometimes the answer is right in front of you all the time. I sure wish I'd thought of this great breakthrough in compression technology!!! :lol:

[T-Will]

LOL!!!!!! :D

[T-Will]

LOL! I'm still laughing! :D Why didn't we all think of this? Duh, this is obviously how the compression technology works!

[Ed Hansberry]

Just imagine that this is video and that's how it works! Sometimes the answer is right in front of you all the time. I sure wish I'd thought of this great breakthrough in compression technology!!! :lol:
Then you could take the the foam and run it through that Ron Pepeil (sp?) thing that sucks the air out so food lasts for years instead of weeks, which shrinks it even further.

[Steven Cedrone]

Then you could take the the foam and run it through that Ron Pepeil (sp?) thing that sucks the air out so food lasts for years instead of weeks, which shrinks it even further.

On the shelf next to the Popeil Pocket Fisherman is the Popeil Pocket PC Compression Algorithm :lol:

Steve

[Will T Smith]

http://www.jacksonville.com/tu-online/stories/050502/met_9322821.html



This is a great link documenting a high tech hoaxer/fraud. In a way though, I somewhat applaud him for seperating money from corporate executives and well to do's who likely didn't earn it in the first place.


Now on to the compression technology.

Would ALL the players be all over it all ready?
No, not necessarily. Technology vendors are notorious for shunning technology when they don't have a "piece of the action". Often, money fights over royalty fees hold up adoption of technology. Good examples of this are, intermittent wipers, wheeled coolers, and more recently ... Firewire. Often, big corporate types try to wear down inventors into selling out for proverbeal peanuts (compared to the value of the invention) rather than licensing at reasonable fees.

Is the concept of Vector/Spline based video formats realistic?
It's absoluetly feasible that video can be described in terms of mathematics instead of pixels. Such a format would significantly reduce bitrates as objects can be "moved" instead of repainted completely.

An excellent approach would be to define colored regions as "blobs" with color points and shading vectors. Blobs can move, reshape, and change shading vectors. Such changes can be expressed as deltas in a video stream. When super-sampled from a 3x video source (2x as dense as DVD), such a format would correct for any precision issues and provide an encoding that is resolution independent. That is resolution is no longer important, only the precision level of the data points describing the objects.

There are some caveats however. Encoding would be VERY, VERY CPU intensive. Decoding would also be very CPU intensive as it would be similar to doing 3d rendering. It would be VERY MATH INTENSIVE.


Are these bit rates realistic?

No way. Even using an alternative method of "describing" a frame. Your average frame contains LOTS of detail that would be difficult to describe in such a small space.

In any case, this fellow is suspect. But I will not call it a fraud without further information.

[denivan]

http://www.jacksonville.com/tu-online/stories/050502/met_9322821.html


Didn't I post this link in the same thread?

Technology vendors are notorious for shunning technology when they don't have a "piece of the action


Yeah right, so Nokia will make a deal with crappy realmedia, but not with something like this, sure.


It's absoluetly feasible that video can be described in terms of mathematics instead of pixels


Yeah, flash is based on that. But to get FMW video based on maths....on a nokia 9210 ? Hell, that thing can't even display over 1000 contacts without crashing.


In any case, this fellow is suspect. But I will not call it a fraud without further information.

I will ;) Look, I checked this guy out when his was posting things on web forumes. His spelling and grammar are so awfull, he writes like a 10 year old kid. I could post the treads, but I'm sure most of you can't read dutch ;)

[bchristian]

The reason why I did post is is because it's very plausible that it is no scam.

I am afraid I have to strongly disagree. It is completely implausible that someone could do what he claims. Just by running the numbers, the way some people already have, can tell you this. Add to that the fact that he has been working on this for a while with no proof still of his claim. I know it might be a little embarassing to you since you posted it if it turns out to NOT be true, but it could also be embarassing to me if it turns out to be true. However, I will go out on a limb and say with certainty that this will not work. Even if he can just approach the level of compression that he claims and I am NOT saying that he ever will, it would probably take a supercomputer days and days to interpret a movie using the methods you described. I don't think the technology is even close if it ever will be.

So, I have done it. I have put this out there unashamedly. If I am wrong, I will admit it.

[bchristian]

Is the concept of Vector/Spline based video formats realistic?
It's absoluetly feasible that video can be described in terms of mathematics instead of pixels. Such a format would significantly reduce bitrates as objects can be "moved" instead of repainted completely.

An excellent approach would be to define colored regions as "blobs" with color points and shading vectors. Blobs can move, reshape, and change shading vectors. Such changes can be expressed as deltas in a video stream. When super-sampled from a 3x video source (2x as dense as DVD), such a format would correct for any precision issues and provide an encoding that is resolution independent. That is resolution is no longer important, only the precision level of the data points describing the objects.

---snip---

In any case, this fellow is suspect. But I will not call it a fraud without further information.

Okay. I will call him a fraud then, if you will not. ;)
I do some 3D graphics and let me tell you, even without animation those files are BIG, much bigger than an MPEG4 file for the frame size. Even adding in vector/spline based animation makes it pretty big. With this you don't have any detail like the real thing though either unless you bring in bitmap textured sufaces or really good shaders. Plus converting video to something like this is not feasible using current technology. Notice I didn't say "not possible". Then again, you could also have built the Hoover Dam by mixing conrete in a wheel barrel. Could you do it?... Maybe. Would it be feasible? No.