| Author | Thread |
|
|
01/17/2004 04:51:33 AM · #1 |
Can someone please settle an argument brewing in my workplace.
On television programs such as CSI, we see criminal investigators remove the blurring or blackout on a subjects face hidden from recognition.
My friend says that this is just televisions portrail of digital imaging.I think that it should be possible from the information held in a digital image to reconstruct the original.
Please help before fists fly. Also a link to the software would be handy because I know he'll not believe me.
|
|
|
|
01/17/2004 04:54:15 AM · #2 |
A lot of TV shows use fake editing, for example blowing up a tiny part of a photo really large, at perfect quality. You can use Photoshop CS to easily show the detail in shadows, but blurring (if you mean motion blur) is impossible to fix with editing afaik.
|
|
|
|
01/17/2004 05:25:06 AM · #3 |
One thing of note is that even with the most sophisticated optics and high powered software the US govt can't reliably resolve the faces of people on the ground from their spy satellites that are designed for those types of "zoom and resolve" jobs. Obviously several other factors play into why a powerful individual or organization might not be able to positively identify someone from an angle of high incidence and through the atmospheric blur but the principle is the same: if your image doesn't have that much resolution then it just doesn't carry the data you need to zoom in a factor of 10-100x and still resolve a face or license plate or anything else into an intelligible image.
Predictive resolution is a different kettle of fish wherein the software would present possible license plate numbers or possible photos based on heuristics that, in turn, came from the cumulative experience of many seasoned intelligence people.
If my memory serves correctly, back when Avid video was coming out (they used to make turnkey video solutions for wedding photog/videogs among others). They sold their top of the line system to a couple of medium to large metropolitan police groups for identification. The story that I remember reading in some literature about their system was that Miami-Dade Co used their system to watch for known perps in large crowds. At the time I installed a couple of their Mom-n-Pop turnkey systems and I remember that the resolution back then wasn't great. I'm sure that with today's optics and software what you're talking about can be performed but only at a close level where faces are moderately distinct to begin with. I don't think anyone but the most well-equipped system can capture motion or stills from a couple of miles away and resolve those images into the same crisp, clear and distinct image that most of us see on TV shows.
Of course, that's all just my 2 cents worth (and may actually be worth less than that with this market).
|
|
|
|
01/17/2004 05:49:23 AM · #4 |
You are talking, I beleive, about the intentional blurring out of a face, and whether that can be reversed. The answer is, "it depends".
IF you know HOW the blurring was done, then a technique called deconvolution may recover some detail. The severe blurring used to obscure facial features on video would be difficult to reverse, and the process would be very computationally intensive, however I would not put it in the realm of "impossible", just "improbable".
There are programs out there that will allow you to recover some detail lost to out-of-focus condition or motion blur. One example is Focus Magic. I've tried this program and the results are impressive, though it can't work miracles.
|
|
|
|
01/17/2004 05:53:15 AM · #5 |
Thanks for the link to Focus Magic, it looks very useful
|
|
|
|
01/17/2004 06:19:29 AM · #6 |
Originally posted by Konador:
Thanks for the link to Focus Magic, it looks very useful |
A bit more on focus Magic; I tried the very limited demo (only allows 5 repeats before registation), and was sufficiently impressed to purchase it. It works near miracles with those "just slightly soft" areas that you wish had been razor sharp.
Like any detail recovery technique, it does raise noise, so the source image had better be very smooth to start with.
|
|
|
|
01/17/2004 09:12:53 AM · #7 |
Many thanks to all that replied so quickly to my question. The link to Focus Magic is useful and the software is now on my birthday list.All the wife has to do now is come up with those six winning balls and the Lamborgini,Mansion,Laptop,Camera and new boxershorts are all mine..............
|
|
|
|
01/17/2004 06:28:24 PM · #8 |
I think I know exactly what Wanderer is talking about here because I am a huge CSI fan. He is talking about when they show a video or picture of a person and the face of that person in the original data is like maybe 5 pixels and no other detail. Then they somehow are able to take those 5 pixels and create a perfectly clear picture of a persons face.
I believe the answer to your question is NO. If the data isn't there the data isn't there. It's all a TV trick.
Hope this helps settle your argument.
Anyone else want to back me up here?
|
|
