Answering my own questions since 2001

Can You Clean That Up a Little?

As a frequent Photoshop user and skeptic, there’s a moment in every crime movie that makes me laugh. It’s when the grizzled detective is back at the office, and asks the local pencil-necked, glasses-wearing lab technician, to “bring up the footage of the robbery”. The lab tech types away at his PC, runs the footage, and our hero says “stop it right there.” More typing, and the frame freezes.

And why does he always type? Anybody who’s ever actually used a computer knows that most, if not all, image navigation and manipulation is done with a mouse or other fancy input device.

Then comes the inevitable question: “can you enhance that section?” The nerd at the controls types some more, and zooms in on the indicated section (often with a box and flashy zoom animation). The zoomed image starts as pixelated, but then slowly comes clear to reveal the license plate, tattoo or coffee cup that fingers the killer.

As far as I know, this approach is technically impossible. There’s no way to magically add more pixels to an image that will make it clearer. What you see is, indeed, what you get. Plus, anybody who’s seen security camera footage knows that the feed is typically, at best, mediocre.

To illustrate my point, I did a search and found this nice big screen-cap from a security camera. Ignore the guy in the centre with the goofy hat. Let’s say you wanted to get a closer look at the guy with the tie and glasses. Here’s what he looks like, zoomed to 400%.

As you can see, there’s no new information–we could tell everything about him from the original screen capture. I applied my limited Photoshop acumen to this image, and I could make things slightly clearer, but this image really doesn’t have much more to tell us.

Is that a beard, or a shadow under his chin? What shape is his nose?

It also kills me that police labs in the movies and on TV seem to have customized software for managing and viewing images. The reality is far more mundane. I saw a news piece recently on the Internet team of the RCMP’s child pornography division. They were using Photoshop to examine images and videos to look for clues.

Just as everyone in the movies still owns an answering machine, this is a minor suspension of disbelief that we accept. I was reminded of this phenomenon while watching The Bourne Identity again, where Treadstone’s entirely pursuit of Bourne depends on their ‘cleaning up’ a feed from a security camera to spot Franka Potente’s license plate. Another movie where this kind of photo-manipulation was crucial to the plot was Rising Sun.

15 Responses to “Can You Clean That Up a Little?”

  1. Chris

    Yeah, this is also a pet peeve of mine – most tech-types probably share it.

    Another movie pet peeve is actors who have no musical knowledge at all playing musicians. It pains me to watch *star-of-the-week* pretending to play piano when they have no idea what they’re doing. If the start doesn’t know how to play piano, but needs to for a critical plot turn or something, at least spare us the close-ups of their hands. *shudder*

    /music snob

  2. Everything Is In/Flux

    Um, what is that license place?

    If you’ve watched movies or television shows with me then you might know that it irks me to no end how the forensics labs have amazing photo manipulation tools that inevitably reveal the truth about a crime or person’s identity. This happens most ti…

  3. Olaf Gradin

    Watch “The Net.” That movie is almost as funny (in retrospect) as “Office Space.” I can’t even begin to tell you all the follies there.

    I might also note the humor I see in the movies showing high-tech legal authorities using sophisticated technology to hold press conferences about some new villan. Who hires people to create visually stunning, interactive DVD content for profile presentations? Seriously, I’m willing to relocate.

  4. Jeff Running

    I’ve also always been fascinated by the simplicity of systems used in movies. Everything has a slick graphical interface with little pictures of what’s going on. I’ve always assumed that this is for the benefit of the general public; whom would fall asleep if presented with the details. Instead of having the sidekick explain what happened the graphics seem to fit that role.

  5. Arild Jensen

    A while a go I saw how they managed to take a fuzzy video image of a car and find out which make and model it was. Not by image manipulation like in the movies, but by using the exact same camera under the same conditions filming different cars. Next they simply compared the fuzzy images to the first fuzzy image. It turned out to be quite accurate.

    I thought that was a very clever way of doing it.

  6. Darren

    ‘Nee: Thanks for that…it’s interesting. That package appears to be essentially an add-on for an Avid system. Avid produces, as far as I know, the industry-standard video and film editing system.

    Regardless, I don’t think I’ve ever seen this system in a movie, and the UIs I do see in movies look a lot cooler than this.

  7. 'nee

    I have a copy of Avid around here somewhere but I can’t run it because I don’t have a quadruple-processor Cray sitting around. Christ, that program takes a lot of juice! And no, you’re right, the movies do make it look a teensy bit easier than it actually is :) But they can do SOME of what the movies claim, anyway.

  8. alison

    this is one of my pet peeves as well. I love the show CSI, but they do this very regularly and it drives me crazy. more then ever because it teaches the general public that this is possible, and that it’s ok to provide low quality source materials.

    i took a look at your example ‘nee, and does help to zoom in on moving pictures, but on tv they seem to be able to enhance it to a crystal clear image.

    it makes me wonder what sort of license they are taking with the other forensic sciences they demonstrate.

  9. Sue

    This zooming in business is just a symptom of the whole “Hollywood OS” problem that plagues movies. For example, Hackers. Watching a bunch of teenagers (albeit HOT teenagers) sitting down punching commands into a Unix prompt is going to be horribly dull, so they created a massive graphically-intensive OS to simulate the systems being hacked. It was a total joke. I’ve actually started watching for movies that don’t use Hollywood OS just to give my kudos to directors who value a bit of reality. “Office Space” was amusing for this – I’m not sure but I think they mixed their PCs and Macs in ways that probably wouldn’t work. Aside from that, the screen shots were pretty realistic.

  10. Ali

    You’re right, what they show in the movies is totally bogus. But what you just did was no better. The ‘zoom’ (in quotes because it is not really a zoom that photoshop gives you) destroys any hope of finding information about the picture. What photoshop calls a zoom is a zoom followed by their best attempt at a blur without telling you so. Why? because that’s what people want out of a zoom of a picture they took.

    If you’re really after information from a photograph, you want the raw pixels to look at, not a blurred zoomed-in version. A picture taken with a real camera is a radial-blur (due to the lenz) of a [slight] guassian distorted image (best approx of the opaqueness of the air) of an interapolation of the real image, only out of focus.

    That’s 4 matrices, three of which have known dimensions (the first three) and the last one depends on the apature and the lens. The middle two have linear entries, the first has sin and cos in it (of linear values, thankfully) and the last depends only on the resolution. So, if you know your camera (i.e. know an approximation to the first, second and fourth matrices) you can compute a decent approximation of the real image back from the obtained image. But that’s way too hard to figure out when you have a bad guy to catch in a movie.

    Thankfully, a second-order approximation to all those values can be obtained quickly (think in real time at 30 fps) if you have a sequence of images from the same camera. You identify an object you know in the scene (corner of a desk or something like that, wouldn’t be too hard to find in security camera footage of a store) and by computing its position and real dimensions in several frames, you can find the constants and expressions you need.

  11. Jan Karlsbjerg

    In Bladerunner, Deckard uses a computer (through a voice interface) to “enhance” a part of a surveilance picture… to the point of showing a face that was hidden behind a column or wall or something. Now that’s enhancement.

Comments are closed.