anyone familiar with this technology ?

[shlezman]

These guys claim 460% improvement over MPEG-4 :
http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104&STORY=/www/story/03-28-2006/0004328087&EDATE=
http://www.eucliddiscoveries.com/

Of course these kind of statements appear from time to time in the video coding field, and usually turn out to be false. I can recall some cases.

It's video object coding but that's about all I could gather. I just wonder if anyone knows more.

[foxyshadis]

It'd be more interesting if they bothered to include a single screenshot with their earthshattering announcement. It sounds as if it's based on wavelets and curve modeling though, which isn't much different than ultra-low-bandwidth snow. Which likely means encoding and decoding is much slower than AVC.

It sounds like the sort of claims real's been making for years, which turn out to be based on 160x120 video that are too damaged to make out. =p Hopefully it's a little better than that.

[Sirber]

@foxyshadis

Any link on "ultra-low-bandwidth snow"? Last time I checked it needed more bandwith than xvid to be watchable.

[foxyshadis]

I was wrong about snow, I've been checking it out this morning, heh. It was some other codec with wavelets and probably combined dct. Maybe dirac? I don't know much about it, maybe it was an experimental codec, since it was terribly slow. (I've got to keep these kind of references handy.)

[Sharktooth]

These guys claim 460% improvement over MPEG-4 :
http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104&STORY=/www/story/03-28-2006/0004328087&EDATE=
http://www.eucliddiscoveries.com/

Of course these kind of statements appear from time to time in the video coding field, and usually turn out to be false. I can recall some cases.

It's video object coding but that's about all I could gather. I just wonder if anyone knows more.
sounds like usual BS... ;)

[Niddhogg]

“ Because we’re building off the existing MPEG-4 standard extensions, we have video compression technology that will eventually work with the hardware and software that is available now,” said Richard Wingard, CEO of Euclid Discoveries.
Motion to move thread to MPEG-4 AVC ? :o

EuclidVision ’s technology is object-based. In simple terms, EuclidVision recognizes objects in the video, like a face, and applies new compression techniques to those objects differing from the background. Current video compression using Discrete Cosine Transform does not look at objects, it just applies a constant rate of compression to the entire frame or picture.
Now if it was true object based encoding...
But nice name for the company, wisely chosen. :)

[PaulC010]

Not sure it was a wise choice for company name. Euclid was good at what he did whereas I predict that this may melt like snow.....

[shlezman]

I tend to agree with sharktooth

[foxyshadis]

Well, by converting a movie into a shockwave flash type format, decomposing into 2d or 3d primitives, you could really compress it hundreds of times more than normal video. The problem is that playback processing requirement rises humongously with the amount of detail. (And the effort of converting it at all.) How would object-based movement help you store someone turning around, or a camera panning around someone, short of building a complete 3d model of them and the scene and rendering on playback?

It'd be a useful technique for old-style animation (though direct conversion to flash is probably easier and much more compatible), I just can't see it being useful anywhere else.

Maybe they use use bigger blocks and really aggressive AQ though. =p

[bond]

i didnt read the thread, i didnt read the news article or whatever it is, i just read "These guys claim 460% improvement over MPEG-4" and i know its fake :D

[mod]

I can reach 461 easily :p

hey, trust me!!

[Sirber]

I heard they use lossless compression ;)

[siddharthagandhi]

Who knows? Hopefully one day there will be a miracle codec that will do HD quality videos at 10 mb a minute. AVC is getting there, but no cigar.

[Niddhogg]

Never say never, one other option, apart from true geometric detection, could be to leave more info handled on decoder's side.
DCT tables perhaps.

[davidlt]

Euclid Achieves 460% Improvement Over MPEG-4 With No Loss Of Visual Quality, Paving The Way For Video Conferencing On Mobile Phones

Concord, Massachusetts (March 28, 2006)- Euclid Discoveries (www.eucliddiscoveries.com), a video compression firm, announced that its technology has achieved compression ratios of 15,168 to 1 for certain videos. The technology, called EuclidVision™, greatly exceeds the current standard for digital video with a 460 percent improvement over MPEG-4, which implies more than a 600 percent improvement over DVD video format MPEG-2 for certain videos.

EuclidVision paves the way for applications like video conferencing on mobile phones, and streaming live, high-quality video to portable devices like PDAs.

“Because we’re building off the existing MPEG-4 standard extensions, we have video compression technology that will eventually work with the hardware and software that is available now,” said Richard Wingard, CEO of Euclid Discoveries. “Bandwidth limitations have stifled the promise of wireless digital video for years. By providing an order of magnitude increase, we just solved that problem.”

EuclidVision uses a new generation of video compression known as “Object-Based Compression” or “OBC,” which refers to technology that analyzes shapes in the video to achieve higher compression ratios. This is a major departure from other compression technologies, including MPEG-4, which are based on “Discrete Cosine Transform” or “DCT.”

The architects of MPEG-4 technology envisioned OBC as the future for this video standard but, before Euclid Discoveries, no firm managed to make it work. The MPEG-4 standard anticipates object-based compression through “Object Planes,” facial modeling, and 3D object modeling – providing only a definition of these concepts without providing the means of employing them toward the goal of high compression ratios.

“While this is a revolutionary concept for the general public, the video industry saw this coming,” said Euclid Discoveries President Bob Werner. “Still, no one thought it would be coming so soon. Many thought this level of mathematical modeling would have taken another ten years to develop.”

Applications
EuclidVision testing has focused on what the firm calls “streaming commentator” applications. This is video that shows the head and shoulders of the subject, which makes the current generation of the technology optimized for video conferencing applications, or simple newscasts.

For streaming commentator video, Euclid can reduce a 23 MB video file to a 1,519 byte file – effectively enabling sub-4Kbps, low-bandwidth streams for wired and wireless applications. The streaming commentator application provides a rigorous proving ground for EuclidVision.

In the coming months, Euclid Discoveries expects to complete tests demonstrating its ability to process all video types including full-length movies, and at increased compression rates. Ultimately, EuclidVision should be able to reduce the current MPEG-4 attainable 700 MB file size for 2-hour long videos down to 50MB – finally making feature length movies as “swappable” as MP3s.

“The potential of object-based technology is so great that it seems counterintuitive to what most people know about compression,” said Euclid Discoveries’ Chief Software Architect Chuck Pace. “By bringing Computer Vision algorithms into the video compression world we’ve taken this first, momentous leap, and anticipate even larger reductions in video file sizes are soon to follow.”

Validation
The company validated the technology by making a head-to-head comparison between EuclidVision and MPEG-4 Part 10 AVC (a.k.a. H.264) – the state-of-the-art for video conferencing – on a network with 9 kilobits per second of bandwidth. This bandwidth-constrained network simulates what people could get from a typical mobile phone connection.

Euclid Discoveries started with a baseline test to show that it could achieve similar digital output to MPEG-4 in this bandwidth environment. Then it lowered the bandwidth to 4 kilobits per second, resulting in a total breakdown of the MPEG-4 output, but a clear, recognizable video using EuclidVision.

EuclidVision compressed a 23 MB 30fps reference video down to 1,519 bytes (15,168:1 compression ratio, 3.56 Kbps bandwidth). MPEG-4 H.264 compressed the same video at similar quality to 8,518 bytes (2,705:1 compression ratio, 19.96 Kbps). Given that MPEG-4 is 50% more efficient than MPEG-2, the 5.6 times improvement over MPEG-4 represents an implied 740% improvement over MPEG-2.

“We’re producing visually meaningful video clips in a bandwidth constrained environment going down as low as 1.6 kilobit per second. This makes many of the bandwidth constraints for digital video moot,” said Jeff Roberts, Technical Product Manager for Euclid Discoveries. “We basically wanted to see how low we could go; we quickly realized that no one gets close to us in terms of operating in a low-bandwidth environment.”

How It Works
EuclidVision’s technology is object-based. In simple terms, EuclidVision recognizes objects in the video, like a face, and applies new compression techniques to those objects differing from the background. Current video compression using Discrete Cosine Transform does not look at objects, it just applies a constant rate of compression to the entire frame or picture.

As EuclidVision becomes more sophisticated, it will be able to identify more and smaller objects, allowing for higher rates of compression. Already, EuclidVision has achieved compression ratios of up to 38,000 to one in a test setting.

“EuclidVision represents ‘new math’ in the video compression space,” said Euclid Discoveries’ Werner. “It took 15 years to move from MPEG-2 to MPEG-4, which represents a 50% improvement. In a fraction of that time, we’ve gone from 50% to 460%.”

Patent Protection There is no company that is solely dedicated to making such effective use of this methodology besides Euclid Discoveries, and the company has effectively shut the door on future competitors because of its diligent patent filings covering key aspects of this underlying technology.

As of December 2005, Euclid Discoveries filed 15 patent filings covering “Apparatus and Method for Processing Video Data.” The patents stake the Euclid Discoveries’ claim to key enabling discoveries in computer vision and image understanding.

“Our U.S. and international patent is the insurance policy covering Euclid Discoveries’ intellectual property, the key assets of the firm,” said Euclid Discoveries’ Wingard. “These protections provide a major barrier to entry for companies looking to move into this area of compression, allowing us to effectively own this space.”

About Euclid Discoveries
Euclid Discoveries researches and develops next-generation video processing and compression technology. In order to meet consumer demand for high quality video on any video-enabled device, the company developed EuclidVision™. EuclidVision is a proprietary technology, which will increase video compression ratios by an order of magnitude - reducing file size dramatically and removing existing bandwidth constraints for the transmission of high quality video. This technology has clear potential applications in a host of industries beyond consumer electronics and telecommunications including surveillance, and image mining. Euclid Discoveries has also developed a rigorous testing environment, EuclidStudio™, to evaluate video compression and quality for MPEG-4 and other compression technologies. For more information on Euclid Discoveries and its offerings, please visit

[mod]

The link is wrong, there's a ")".. :)
BTW, another thread about this was started a few days ago. They don't show any comparison..

[Sirber]

This testing environment, called EuclidStudio™, is available for download from the company's website.No download in their site.

[mgh]

download at
http://www.eucliddiscoveries.com/information.php
asks phone number among other details

[davidlt]

Link: http://www.eucliddiscoveries.com/download/euclidStudioDemo.zip
Size: ~5.5MB

[mod]

thanks davidlt :)

EDIT: it doesn't work, I followed all installation steps but always get app error.

[CEC]

So, is it good???
Better than AVC???

[davidlt]

Unpacked, about 13MB
But it didn`t work for me, I need reinstall windows, but can`t at this moment.
It crash all times, than I start it :( only after few days I`ll be able to reinstall windows.

P.S. Read readme.pdf, for more infomation, who to install it...

Modes
The demo version of EuclidStudio™ has three basic modes:
Encode/Decode
Select an input file of type .avi, .cif, .qcif, or .erv and press Play or Single Frame to feed the video frames to
the encoder. Depending on the settings you made on the Control tab, most, if not all, of the visualization
windows will be updated while the studio runs.
Decode Only
Select an input file of type .264 and press Play or Single Frame to feed the encoded bitstream to the
decoder. In this mode, only the “decoded” visualization in the Viz Stack window will be updated.
Record raw video
To record a raw video from the camera, or image index buffer, select Camera as the input source, select a
filename for the output file, or generate a default, then toggle the Record button on. To record live video
from the camera, click the Enable Recording button and press Play to begin the recording. To record from
the buffer, check the appropriate checkbox and press Start Recording.

Maybe someone could make some small videos for example?.. I would like to see if it is true better than MPEG-4...

[Zarxrax]

From looking at the program, its obvious that this encoder is not for the use of encoding "real" types of content. Maybe more for video conferencing or something.

The included test clip is a 176x144 uncompressed avi at 1 frame per second. It is 1.45mb. The content is a news woman and she hardly moves. After running it through the encoder, I got a file that was 3.76kb.

I encoded the same sample clip to xvid, at constant quant 4 plus b-frames. The output came to 16kb, and both files appeared to have roughly the same quality.

The same video encoded to x264 through the video for windows codec (build dated december 2005) at default settings resulted in a 14kb file.

So ok, if you use a very specific test case, this thing performs far better than xvid, LOL.

I was encoding a 100 frame sample of "real" content, but it crashed partway through.


Update: A friend who actually knows what he is doing with x264 made a 3.61kb encode from it that looks just as good as the output from the EuclidStudio codec, in my opinion. So even in this test case they provided, normal h264 seems just as good!

[Zero1]

Yeah, you should have seen the "sample" material. Absolute joke. 20 seconds of virtually NO motion @ 1fps. What the hell is that all about?

The config files looked like the ones I see in the reference software (but I've never been able to compile it). Do the differences stop there or maybe it's just based on (or even just is) the reference software? I wonder this cause getting the file down to 3.61 wasn't much of a task

[bratao]

When i read this some thime ago i think
"hey, its like a hack to transform video to flash way. it never work for complex movies like matrix"
I was right

[foxyshadis]

Someone needs to send same-size xvid and x264 files along with a press release to all the news organizations that have breathlessly picked up the story. Perhaps even add some hyperbole about xvid and x264 being the new wave of collectively developed software pushing standards to their limits and defeating commercial solutions. The Web 2.0 flavor almost guarantees publication. :p

[Doom9]

Whenever you hear marketing people talk, you have to activate the noise filter ;)

[Sharktooth]

Whenever you hear marketing people talk, you have to activate the noise filter ;)
you mean the NOSE filter (for the smell of BS...) :D

[bratao]

I read the whole site , and its clear bullshit
Whey using mpeg4-avc ! plus their "Object-Based Compression”

But they talking that can get 50 megas for a movie, its BS i doubly if their "Object-Based Compression” can do anything with a full action movie, like matrix...

[Coderjoe]

I read the press release when it was posted on wednesday (http://forum.doom9.org/showthread.php?t=109322) and knew it was bullshit. Object-based compression has been part of the mpeg-4 spec for years. It just doesn't see much use because you have to have the objects separated out when you encode. Their press release also says "for certain videos" and that their software is optimized for video telephony, or in other words, "talking heads". I just didn't know how much bullshit there was until Zarxrax and Zero1 mentioned their sample clip. Zero1 also tried compressing a clip from a Gundam show, and it came out all weird.

(I tried commenting on Wednesday last week, but was bit by that stupid "must be registered 5 days before posting" message. :angry: )

[Zero1]

Here is said weird Gundam clip...

http://aflux.deltaanime.net/Zero1/MP4/euclid.mp4

Now the input was 704x480 @ 23.976 fps lagarith. The resolution it output is obviously for mobile phones and the likes, and the fact it decimates the framerate also suggests the same.

So is this really object based coding, or is object based coding referring to the fact that their software crops and resizes an input frame to that parts that are most interesting...

Also, AFAIK H.264 does not have object based coding since it's specifically aimed at natural video coding, as opposed to MPEG-4 Visual which does have OBC.

[Coderjoe]

Well, as expected, the various news people are falling for this bullshit (http://www.boston.com/business/globe/articles/2006/04/10/firm_squeezes_films_into_a_download/).

[siddharthagandhi]

How do I get it to work...I downloaded everything i dunno how to use it???

[Wolfman]

April fools or stock scam?

[siddharthagandhi]

yo can someone tell me how to use it? (infirma im talking about)

[foxyshadis]

But... this isn't the Infima thread at all. Infima is (basically) something like winrar wrapped into a codec wrapped in a thousand license violations.

As for this codec, well, it seems more prone to crashing than anything, maybe you need a specific system setup (don't bet on the vaporware company supporting it).

[Leo 69]

It is a very old program, BTW. Have a look at the date of the files, which are included in the archive and also at the splash screen, when you execute the program. It is all from the year 2004.
Мore than that, the program doesn't work at all. I can see only an error message after a splash screen.

[Soulhunter]

New revolutionary compression technique known as PNG produces 10x smaller files than h.264 lossless *cough* for this (http://img160.imageshack.us/img160/3778/1114zu.gif) sample *cough*


Bye