[G��g]

Wait! I found more!



Sorry pisangkecil, but loosely speaking, your friend has been blowing sh*t up your ass.... aahhh.. the wonders of Photoshop.

[aNoah]

You guys don't know nothing.

I have photoshop 7.0, and it enchances images. Yesterday I took some pictures from the UFO show, and then I enchanced the images. You'll never believe the results. I saw Bush and Gore making deals with the aliens. They were talking about how they were gonna get their votes really close, and when US is in a state of confusion they were gonna take over the world with the aliens. Yes, Photoshop7.0 is amazing isn't it.

-ok, enough crap out of you aNoah...
-Sniff, bye

[Chapel]

I'll tell you this much. It is not possible with the picture formats available to the public.

[This message has been edited by Chapel (edited November 23, 2000).]

[Plouffe]

I dont know if anyone gonna read this BUT actually image enhancers do existe. My father works for the canadian air force and at the school he teaches in they have VERY VERY VERy expensive photo software. Just the image enhancer cost the government about 1 million dollars NO JOKE , im totally serious. these things are not available to the public but you would be suprise what the goverment get their hands onto. Laugh all you want but im willing to bet withing the next 5-10 years this software will come for public use.

[Chapel]

I'm pretty sure that is what I just said.

I work for Lockheed Martin.

[FatPenguin]

Singa: sure you can make it look nicer, but you cannot get any more information out of it. Similarly you could just paint directly over the pixelated eye, and end up with a nice image.

you cannot take a bunch of pixels, put them through some software, and then get more information out of the result than the origional pixels. This is not possible! In the example of the car license plate, the software would have the exact same data as you do, which meens it has to be better at 'guessing' than you.

there really isn't anything else to it.

[RoadMaster]

to make a long story short it's the hardware. There is "photo enhancing tools" but those are called video lunchboxes usually, and they work with photographs and videos, not digital captures, since they are attained in different methods, the video lunchbox can zoom in on a video clip, but can't get details of a pixel. Videos and photos don't use pixels if you haven't noticed

------------------
The Elves, EVIL sinister elves...

[pisangkecil]

Well I'll think next time... I'm much to gullible

[This message has been edited by pisangkecil (edited November 24, 2000).]

[Singa]

FatPenguin:
No you can't add 'extra' info, but you can use the available information better that resizing and smoothing it bilinear. That's what I meant; There is more info in such an image than than we can see.

And I am convinced that resizers with a very soffisticated AI can add extra info aswell... For instance: I know what a human eye looks like, so if I have a very low-res eye image, I can add detail after resizing. Give that knowledge to a computer and he can do it aswell (if he has good enough AI).
I admit that there might be a little guessing involved.

As for low res (unreadable, but not too low-res) images of license plates, I'm sure it's possible to determain the number to a certain extent when the computer has a database available with all existing license plates.

[Fruitbat]

A image compressed to one single pixel will not yield any additional data no matter how well analized. However, a pixelated photo has a LOT more data than just what we see.


Very Simple Example (V.S.E.):
Between two different colored surfaces in a picture, the computer has to pick a soft in-between color to make a nice transition. But notice those in-between colors are all different. By noticing how the in-between colors are closer or farther away from the original solid color you could actually re-construct the original vector shapes. The fundemental algorithm is the same for reconstructing a photo. A photo is much more complex but it is the same process and I would bet any amount of money that license plate could be rendered readable with the proper software.


It really shouldn't be any surprise that this is possible. Your mind does this all the time without any help from software, it just happens automatically. If you can recongnize the guy in the photo above you have already done it.

[jayce]

Okay i'am like the only dutch person to view this thread??

Andr� O. Zodom translated from dutch to english reads something like :

Andr� o soStupid ;-)

Looks to me like a succesfull april fool in november..

[FatPenguin]

-Originally posted by Fruitbat:
...Your mind does this all the time without any help from software, it just happens automatically. If you can recongnize the guy in the photo above you have already done it...


Exactly. But to suggest that a computer can do it better than a human is ridiculous. Yes a computer could figure otu that that first example should be shown as a straight line, but so can any person. You recognize the second image immediately as abraham lincoln, which is more than any computer is able to do. You are talking about artificial image recognition being better than that of natural image recognition. Ai just isn't there yet, and i don't believe for a second that it will be in the next century

[Singa]

I don't think AI needs to be very high level to be able to recognise things in images like we do, With some simple rules/algorithms (or whatever) and a large (internet based) database, computers sould be able to recognise and enhance quite a lot...

[Fruitbat]

Fat penguin, the point isn't that the human eye can't aproximate the colors in my example (obviously it can), but that a computer could figure out the EXACT angle of that that line which even a brain wouldn't do properly.

Obviously if a picture that has random colors, and you compress those down, then you can't reconstruct them back, but a photo is not random at all. It has obvious patterns (lines, shadows, highlights, colors) that follow consistant rules. Of course any program that disects this information and re-packages it as a sharper image will be an approximation, but it only needs to go as far as to make the image more readable, not create an infinate zoom.

This is a difficult problem, but not half as difficult as putting a guy on the moon. You can shake your head, but some guy at the CIA is re-sharpening a photo just like this as we speak.

[SizzorIntake]

WTF is that thing in the window???

it looks like an ass!

[FatPenguin]

'You can shake your head, but some guy at the CIA is re-sharpening a photo just like this as we speak.'

I'm not talkig about pohotos. Individual photons hitting a photographic plate is completely different from a bunch of pixels

'This is a difficult problem, but not half as difficult as putting a guy on the moon.'

Absolutely not. It is an infinately more difficult problem. I'm sure there are some specific, semi-relavant situations in which computers can be used to greatly accelrate the process of whatever, but as far as straight image recognition goes, computers are just not capable of aproaching the human brain

[dr . bang]

Image enhancer actually works! I have a proof

[Starseed]

People, I read his post, and I laughed too . . .

BUT,

Envision a young person, who's seen the photo enhancement shot a million times on tv and in the movies. Envision a kid with no experience in photoshop or another image manip program. Without the concept of pixels, and the effects of sizing up (losing quality) and sizing down (no loss) . . . are you really surprised this kid, and a million others would ask the same thing with a straight face?

-mt

[Fruitbat]

I'm not knocking the human brain, it's a great thing, but there are so many things that comptuer can do that a brain can't (or hasn't been trained for). Sequence DNA, plot paths for ballistic missles, simulate nuclear explosions, help solve Fermates last Therom by brute force...ect. Just because the software isn't avalible in a shareware version doesn't mean it doesn't exist.

Here is an example of algorithm used to give detail to a fuzzy image: http://www.adires.com/~castleman/lin_com1.html

And here is an example of two very pixelated images combined with the help of a computer to give an accuarate image: http://ic-www.arc.nasa.gov/ic/projects/bayes-group/super-res/2d/

Here is an artical of image enhancement software used in video: http://www.science.nasa.gov/headlines/y2000/ast31aug_1.htm

[Paqmann]

hehe me 2 me 2

ORIGINAL:



MY VERSION:



This was developed by the space aliens and they took me up and uhmm and they uhmm yah and so i came back and now im the smartest man alive and i figgered it out.

bwahaha its called the MOSAIC command in PS. funfun. I would like to mention that i work at a company that makes movies and all that fun of stuff (sorry no lockheed martin) and there's a lot of image enhancement, etc. technology for VIDEO. But that's the computer calculating which pixels are the same over MORE THAN ONE frame.

I don't understand how someone can get more out of information. If a pixel has a color value of (125, 125, 125), thats it. there can't be some hidden values, no matter WHAT the rest of the image looks like.

the first guy thats cool, some mean person played a trick on him. is pretty funny tho. But the other guys.. it aint possible.

-matt

[kanabis]

i agree totally with what fatpenguin said in all his posts ....

just read his posts and youll understand

[Liquid!]

http://www.sijun.com/dhabih/ubb/Forum2/HTML/007012.html

Sorry I started a new thread for this, as I obvioulsy shouldn't have. But contrary to popular belief I think the possibility exists.

And as mentioned in that thread the technology for "face recognition" by a computer already exists. And we're not talking about straight on mug-shots, but about video capture of off-angle faces. It's being used at border crossings to facilitate the apprehensions of known fellons... sounds very orwellian I know.

-c

[Fruitbat]

Of course I read the pages. I am aware that they are not the exact answer, I am just pointing out what seem to be similar and interesting technology. Anyway, I am posting on the other thread now.
http://www.sijun.com/dhabih/ubb/Forum2/HTML/007012.html

[Jaymo]

Sorry to go on on this thread, but I wanted to reply to FRUITBAT:
Hey, did you even read the Nasa-Stuff you posted the URLs of? They used 63 slightly different low res scans to produce a higer detailed version. This is somwhat different from trying to do the hires version directly from only ONE lowres, no? And the "Space-variant Image Enhancement" is more of a sophisticated sharpening algorythm; the source can easily be recognised as what it is. Not the case with that plate.
IMHO all the computer could do is specify tons of possibilities and impossibilities concerning the number on that plate. If at all.

[Skyraider]

Chapel,

how come Lockheed Martin don't have a good 3D artist at their service? Their 'artist impressions' of the JSF look pretty crappy for such a high-tech company...

What's your function there, btw.?

[pisangkecil]

Well, anyway... thanks guys for all the info... it got me thinking