What is Digg?62 Comments
- julienbh, on 10/12/2007, -1/+18I won't ask of what kind those videos are...
- duk0r, on 10/12/2007, -1/+18I knew keeping those old .mpegs from my kazaa days would pay off.
- wibblewibble, on 10/12/2007, -9/+25Start by converting the moon landings
That is all. THank you :) - Kielrandor, on 10/12/2007, -4/+16How about a video of crime that has been recorded on one of those "*****-ass cellphone cameras". Critical evidence of guilt or innocence can be recovered from this "waste" of taxpayer dollars.
Or how about a video of some Osama-wannabe spouting off the regular retoric of an upcoming attack. Optimize the video a little bit and, oh, whats that in the window? A street sign? Why, that's Fifth and Broadway! Break out the JDAMS!
Get a clue ya moron. this isn't targeted at fixing your retarded step son's birthday video.
Update: the above was suppoosed to be addressed to Willynilly - creednmd, on 10/12/2007, -0/+11Huh?
Why don't they just use the technology that is on virtually every episode of CSI, where they zoom in on a reflection in someone's pupil and process it to get a clean image of the perp's face, or get the name of the victim by cleaning up the scrawled name on the laundry tag just sticking out of their pocket?
Surely it can't be that hard to apply the same software/algorithms to a moving image?
(And, yes, for those that are hard-of-understanding, the above *is* a joke...) - neurobox, on 10/12/2007, -0/+8Actually, in slow moving footage, there is a great deal of information that can be extracted from neighboring frames and re-added to each frame using motion analysis and compensation. Think of a panning shot of a golf course (see example), but think of each pixel in the original low-res frame as tiny data points with so many "degrees" between them - if you can figure out exactly the sub-pixel motion of the camera, you can plot more points in their places beween any one frame's pixels, painting a more accurate picture of the scene as a whole.
- olegk, on 10/12/2007, -2/+9They remove MPEG artifacts pretty well, but that has been done before. There are no "recovered" details. I don't see what's so impressive.
And "military tech" is just laughable. Cmon, stop BSing. - UnglueD, on 10/12/2007, -0/+5Enhance. Enhance. Enhance...
- pbaehr, on 10/12/2007, -2/+7The sample videos they show are pretty impressive:
http://www.motiondsp.com/OnlineVideoDemos2.html - leoCT, on 10/12/2007, -0/+5putting the videos on youtubue probably does not help much. The second comparison is the best, as you can actually read the book titles.
- controlguy, on 10/12/2007, -0/+4Your right you can't add additional detail to the picture, but you if you had multiple pictures of a similar view in low-res, you can combine the information in all of those shots to make more detailed images of each because, well, there's more information with multiple (similar but not same) shots. There must be differences between the shots, though, for there to be information in one not contained in the other.
This happens all the time in videos, and this algorithm leverages it. That's why they say the shots should be slowly changing, otherwise there is less information it can combine. - notfred, on 10/12/2007, -0/+3I also envisioned it, and I have to attribute the delay to someone less lazy than me not reading my mind.
- irieKEN, on 10/12/2007, -2/+5Oepapel: This is very different than simple up-conversion. This software infers information by examining the difference between pixel positions in sequential frames and constructs a final image to better represents what the camera actually saw. It is quite amazing, but not an original concept. Astronomers use similar technology to derive information from telescope images (Hubble Telescope images are just fuzzy pixel blurs before image processing).
One consideration to keep in mind about this technology is that it is limited to static objects in the real world (camera can move, but objects cannot). - oepapel, on 10/12/2007, -3/+6They already have this. All upconverting DVD's do this exact same "military tech".
And of course there won't be any more information in the image. No amount of processing can add data to video. They are just moving pixels around so that artifacts disappear. This is just a generalization of the 3D comb filter that is found in $10 DVD players.
If the tech was really that good or that unique, the obvious market would be the consumer video electronics manufacturers. I suspect that it is neither. - crawfishsoul, on 10/12/2007, -0/+3@15charmaxwtf
Pink Floyd's Peter Clifton says he forgot to return them. But really NASA is looking for the magnetic data files, not the 16mm that PF claims he has/had.
http://digg.com/space/Pink_Floyd_s_The_Dark_Side_Of_The_Moon_Produces_Missing_Apollo_11_Footage - chaosmachine, on 10/12/2007, -0/+3"hard-of-understanding".. that's brilliant.
- oepapel, on 10/12/2007, -1/+4"Oepapel: This is very different than simple up-conversion. This software infers information by examining the difference between pixel positions in sequential frames and constructs a final image to better represents what the camera actually saw."
That's what an upconverter does as well. That's the difference between an upconverter and a scanconverter. A scanconverter does not use multiple frames; an upconverter does. Now, there are many algorithms for spatio-temporal conversion and this one may or may not be better for a certain class of video but the concept is hardly new. - masamunecyrus, on 10/12/2007, -2/+5@langford:
TV manufacturers already make their own codecs/algorithms/schemes to stretch a 480i image to the HDTV's native resolution.
However, some manufacturers do this half-assedly, so maybe it'd be of interest to them. - 15charmaxwtf, on 10/12/2007, -1/+4There was a story, on here I thought, saying some guy had found them.
- webcrumb, on 10/12/2007, -0/+2Your argument makes no sense. You mean you shouldn't give military funding to research military projects because you don't like the university? Or because they are left-er than you they can't understand military technology? Or because they question the Government they should be shipped off in a boat?
You're why the USA has become America. - cuoops, on 10/12/2007, -0/+2NASA developed something like this before 98. It has helped law enforcement catch some people.
http://www.msfc.nasa.gov/news/news/releases/1999/99-073.html - bludo, on 10/12/2007, -2/+4For example you can read the name of the books much better in the second film
- CiXeL, on 10/12/2007, -1/+3this is probably based on the frame comparison technology, not sure exactly what the name of it is but it works by comparing the differences in frames of video to pull out maximum detail. each frame contains details of the view in front of you and they just pull out all those details. when they remastered gone with the wind using this technology they had to make an ethical decision about digitally cleaning stuff up because the picture was so clear that you could see how they glued or taped stuff up. the film north by northwest was also remastered using this technology.
im just waiting for someone to create a version of it so you can run video captured VHS home videos through it to clean them up to sparkling quality for dumping to dvd. i dont think that will happen for awhile though due to patents. - inactive, on 10/12/2007, -1/+3Quit posting that. People will find it and Digg it on their own. Obviously, if you have to spam it, chances are that it's not worthwhile in the first place, and spamming it will only make it worse.
- nomore, on 10/12/2007, -0/+2Cool. Kind of like, reverse MPEG. One hell of an algorithm.
- ElFredo, on 10/12/2007, -0/+2Anti-Lamenessing Engine (ALE) is an Open Source lib that can enhance the definition of images from several sources:
http://auricle.dyndns.org/ALE/
The 2D gallery is pretty impressive. - inactive, on 10/12/2007, -0/+1This'll make 3G videos actually worth a damn.
- oepapel, on 10/12/2007, -0/+1"Sounds like the same process used in the old "Snappy" video snapshot device. It could generate very high res stills from regular video by interpolating multiple frames."
I remember the Snappy. It was short lived and was the first to die when digital interfaces supplanted analog ones. The software that it came with it worked pretty well, considering the fact that the digitizing circuitry was crap. Despite that, it was successful, albeit short lived. - zweben, on 10/12/2007, -4/+5There is no 'fix' for low resolution. You can't add detail to an image, period.
You can, to a degree, fix compression artifacts, which is what is going on here. The processed video doesn't contain any detail that the original didn't have, but it can make it easier for a person to see. - neilmarkellis, on 10/12/2007, -0/+1I'd sue them, obviously they have access to mind probing technology and stole your vision.
- marioluigi123, on 10/12/2007, -0/+1I think this could at least help YouTube get some (somewhat) better looking videos.
- psyduck, on 10/12/2007, -0/+1I can't believe they did mil research at UCSC without the hippies throwing fecies at them and driving them away.
- dwight0, on 10/12/2007, -0/+1didnt they use this years ago to clear up images from out space to telescopes?
- Darkness123, on 10/12/2007, -0/+1I hope they have this as a cell phone program, so you can record the LQ video and it converts it to as much HQ as it can. Also hopefully if YouTube and Google Video can put this onto there servers so it can increase the quality of the videos on their sites.
- zephc, on 10/12/2007, -0/+1I love santa cruz, i was born and raised there, but I have to admit i was shocked - I think the real news is that anything for the military was done at UCSC.
- Texpatriate, on 10/12/2007, -0/+1Sounds like the same process used in the old "Snappy" video snapshot device. It could generate very high res stills from regular video by interpolating multiple frames.
- oepapel, on 10/12/2007, -0/+1"You very clearly don't have any experience in signal processing."
Really? I happen to have 14 years of experience in image and video processing. They both fall under signal processing. I'll take the pepsi challenge with you any day of the week.
"> The only thing that is unqiue here is the algorithm
Really? Way to go Sherlock. That's kinda the whole point."
Read the original post dumbass. The Original Poster said that he envisioned this process but was surprised that no one had implemented it. I answered that it had indeed been done before. This was just a (possibly) new algorithm.
"> And if it was really that good or unique for a given class of video, the consumer electronics companies would be all over it.
Not if the algorithm in question is far too difficult to implement in electronics."
The current HDDVD and BluRay players have a tremendous amount of processing power. The next round will have even more. Just because you can't do something today doesn't mean that it isn't still worth millions/billions in the future. So my comment stands.
"Let's put things in persepective with an illustrative example. Guess what, DVD players don't support the full MPEG4 spec. Why? Because doing so would require far too much computing power. "
Wrong idiot. DVD players support the MPEG2 spec, not the MPEG4 spec. And they don't support the full MPEG2 spec either.
"Actually, most existing *computers* knowadays can't deal with the full MPEG4 spec in realtime. The algorithms exist, they just can't be put to practical use."
You seem to have confused implementation with specification. A spec can have multiple or even no implementations. And a spec doesn't necessarily imply an algorithm, only a data format. Spec's define and encourage interoperability. They may or may not also suggest an algorithm. But even when they do, an implementor can substitute a different but compatible algorithm. - volatileacid, on 10/12/2007, -0/+1I think alot of people here have over complicated or misunderstood matters. Let me try and break it down if I can, with my understanding of it all.
What this technology does is; process frames (obviously) to improve the quality of an object (i.e. the tank) by taking the best consituents of that particular object from a series of frames to ultimately build the best picture it can. In one frame the image of the tank might not contain enough data for the image to be clear, but after tens of frames, all the data on that object captured will result in enough information to produce a better resolution of picture.
So it may just be that the images start off blurry, but then after a bit of motion, things get clearer as the algorithm is able to build a better picture of the various moving objects in the scene. (Also, even if the objects aren't moving, but the camera is, the effect could possibly be somewhat similar)
Hope that helps! - justice7, on 10/12/2007, -1/+2lol csi
- volatileacid, on 10/12/2007, -0/+1yep, your eye test is definately overdue. At least that's one benefit of all this for you!
- frice, on 10/12/2007, -0/+0This seems more important for *photo* cameras. Just point the camera in the direction you want a photo of and the longer you keep it in that direction, the higher the resolution gets...
- oepapel, on 10/12/2007, -1/+1It has been implemented many times. All DVD players do spatio-temporal correction (as part of the 3D comb filter). This is just an adaptive variety of that. In fact, the newer 3D comb filters as implemented on ATI hardware are adaptive too. It's also done in real time!
This is cool technology but it is not new or unique. If it was, the money would be in selling the algorithms to the consumer electronics manufacturers. - Lycander, on 10/12/2007, -7/+6Or wait until someone finds the lost original film. I hear it's much higher quality than the televised version.
- oepapel, on 10/12/2007, -2/+1"Upconverting DVDs will only use the information found on one frame and make it higher resolution. This technology takes multiple frames of almost the same content to build better frames.
Wrong. You are talking about a scanconverter. Upconverters differ from scanconverters in that they DO look at multiple frames. And ALL DVD players and TV's with a 3D comb filter do spatio-temporal conversion.
"The processing is so slow that no DVD player could do it real time even if they decided to add this to DVD players."
This is laughable. $10 DVD players do this.
There is abolutely nothing new about spatio-temporal conversion. The only thing that is unqiue here is the algorithm. And if it was really that good or unique for a given class of video, the consumer electronics companies would be all over it. - inactive, on 10/12/2007, -3/+2All I have to say is "Hot damn I love technology!"
- idntunknwn, on 10/12/2007, -2/+1@oepapel
You very clearly don't have any experience in signal processing. Please stop commenting, you have no idea what you are talking about.
> The only thing that is unqiue here is the algorithm
Really? Way to go Sherlock. That's kinda the whole point.
> And if it was really that good or unique for a given class of video, the consumer electronics companies would be all over it.
Not if the algorithm in question is far too difficult to implement in electronics.
Let's put things in persepective with an illustrative example. Guess what, DVD players don't support the full MPEG4 spec. Why? Because doing so would require far too much computing power. Actually, most existing *computers* knowadays can't deal with the full MPEG4 spec in realtime. The algorithms exist, they just can't be put to practical use.
And do you want to know just how old the MPEG4 spec is? Much of the work is 20+ years old. - atillatheliger, on 10/12/2007, -2/+1@oepapel
You are wrong on this one. Upconverting DVDs will only use the information found on one frame and make it higher resolution. This technology takes multiple frames of almost the same content to build better frames. The processing is so slow that no DVD player could do it real time even if they decided to add this to DVD players. - brentis, on 10/12/2007, -1/+0The sample images certainly would be easy to produce...
1) Start out with relatively clear image
2) Convert image to jpg at 50% compression
3) Label compressed.jpg "before" and orginal "after"
Where's my angel VC. $500k is rather low though. - iloveliberals, on 10/12/2007, -1/+0Maybe he means that anti-military staff there (if there are any) should refuse helping the military on principle. You know, practice what you preach. Me, I'd have no problem w/ it. Name one federal government expenditure, apart from the military, that I couldn't live without. Why I do believe even Al Gore got military funding to invent teh internets.
- iloveliberals, on 10/12/2007, -1/+0Of course you can add detail to an image. Open the image up in Photoshop, add all the detail you want.
I hear what you're saying, though, and I think you're missing the point. They're not really extracting detail from a single image; they're extracting detail spanning across successive sets of images, like the Hubble telescope example someone mentioned elsewhere.
And yes, you can extract detail from a single blurred image, too. Google for "blind deconvolution" and you'll see excellent examples of just this phenomenon. -
Show 51 - 62 of 62 discussions
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is