Slashdot Mirror
"Dream Team" to Create Gigapixel Photo System
neutron_p writes "An eclectic group of artists and scientists that organizers have dubbed the "dream team" of imaging and visualization are gathered at New York University this week to begin to create a photographic system capable of capturing and displaying a gigapixel of visual information in a single image. The first Big Picture Summit, Dec. 8 and 9, is organized by artist-photographer Clifford Ross. Ross says his goal is to bring closer to reality his desire to create a "you are there" photographic experience for those who have not personally witnessed the sublime beauty of natural scenes such as Mt. Sopris in Colorado."
Check out the grand canyon in gigapixel glory
The human eye can only resolve the equivalent of a couple megapixels, so the lack of "you are there" is not really a fault of image resolution. It's the lack of real depth that is missing from fotos. Stereo photography is a step forward, but it doesn't allow for natural focus changes and good (high res) stereo vision systems are far too expensive.
Other than recreational uses, what else could this be used for? Telescope cameras pop to mind for space imagery capture. I think current systems use very high-resolution cameras, though anything that drives down prices would drive up quality.
Has someone applied Moore's law to digital camera pixel amount?
-Hell hath no fury like that of a woman scorned for
I have personally found even lower-resolution 3D pictures to look much nicer than high-res 2D pictures. Combine something like this with stereoscopic glasses and it would be like "being there". I wonder if Mr. Ross has considered this.
I still think the 2.5 gigapixel photo is the best. The detail is incredible, the photo is interactive, allowing zoom capability. You can zoom all the way in and read license plates and see parking passes. http://www.tpd.tno.nl/smartsite966.html
given eye's optics, you can resolve about 1.5 mega pixels. This is assuming that the picture is kept at a distance, so that it occupies about the same area as 50 mm lens would provide on 35 mm camera (or 3x2 feet picture kept at 5 ft away). This is theoretical limit based on perfect print. Since most photos have some artifacts, you reach saturation at a slightly higher pixel count.
If your monitor is more than 1600x1200 and if you want to do pixel by pixel comparison of two photos on a single monitor (each photo size 800x600), then it is not possible to do so without moving your head.
In order to see 1 giga pixel, you will have to be incredibly close to the photo compared to its size and also will have to move up/down/side to see the details at different places.
Higher magapixel beyond 4-6 MP is only good for cropping, zooming, scientific data etc but is not of much use as a single print, specially if it is to be viewed as a whole.
The primary reason, though, would simply be that photographers are using digital cameras in many places where they work quite well, and they would like them to eventually be the primary workhorse for most photographic needs.
Well, that's a great idea. However, since the conference was ORGANIZED by an artist, wouldn't it be just a tad presumptuous to ditch him and start on your own?
And anyway, this is a natural photographer, probably an artist like Ansel Adams, an artist who "really understands what he's looking at" (to quote Feynman). It's not some hippie "concept"-artist welding shopping carts together (not that there's anything wrong with that).
One of the interesting possibilities for cameras with that much resolution is that photography can become a question of choosing a view of a larger recorded image rather than simply recording that cropped view.
This way you can crop your photos OUTWARDS and not just INWARDS after the fact.
This of course has all kinds of privacy implications too (why shouldn't the photograph be an all round view that includes the photographer?)
Wait... you know of a "particularly strong" workstation that can interactively manipulate a 190Gb image? (4G * 48b)
Storing one copy is one thing... storing multiple working copies and interactively working on them is another thing entirely.
Why? 8x10 cameras have existed for 100 years. Using modern film and a drum scanner will create a digital image with more than 1Gb of pixel data.
That's like saying why bother creating better compression formats when you can already compress a 4 minute sound file to (hypothetical) under a meg at 128kbps quality by encoding it to mp3 128kbps first, then to wma 128kbps. You're doing one lossy conversion, light to film, then a very different type of lossy conversion, scanning film to digital. Sure, it works pretty well in practice, but it's far from optimal.
And many photographers obsess over making things optimal. It's why they buy $3,000+ lenses. And not just one, either.
Anyway, the answer for "why" seems pretty obvious to me, at least.
Random and weird software I've written.
The Gigapxl team have taken a far more scientific and rigorous approach than Ross.
For example, they have:
Both Gigapxl and Ross are using converted 9x18" aerial photography cameras, with vacuum backs, but Gigapxl has taken steps to ensure the maximum performance of their equipment. Ross has not.
And, as far as size goes, there are photographers using cameras up to 20x24.
The "dream team" is really just the "hype team".
Reading Slashdot is ruining my spelling and grammar.
Sure, it works pretty well in practice, but it's far from optimal.
It's a real shame that it only works in practice. Because working in practice doesn't matter at all it has to work in theory. (Incidentally, this actually does work in theory, too.)
Also, your analogy is seriously flawed. It's much more like taking a record (a record with few flaws that runs at a very low speed) and encoding that to an audio format somewhere in the neighborhood of twice the quality of a CD.
The record player you do this with is like the camera. The equipment that originally recorded the record on is like your camera. The microphone is like the lens.
This is also in a world where recording straight to digital doesn't sample fast enough.
As an aside, the reason that digital pictures are harder is that sound happens over a very very long period of time compared to photography. You've probably noticed that most digital cameras absolutely suck at taking action shots in all but the brightest light. This is because the stuff that detects light in the camera doesn't work as fast as the chemicals on the film and doesn't scale as well. The reason that scanning works is that a scanner has an eternity to deal with the image compared to the instant that a camera has.
Help I'm a rock.
I'm a little confused by some of the aritmetic. It looks like you multipled 19 square inches by a linear 2400 dpi.
Also, according to the New York Times piece on this guy, this first prototype is a very large format negative film camera, and the image is scanned from the negative on a drum scanner.
Aside from that, I think people spend too much time worrying about resolution and not enough about bit depth. I forget the exact numbers, but B&W negative film has a contrast range of 10 or 11 orders of magnitude. A B&W print is less. Color slides less, and color prints even less. But even a color print exceeds 24bpp (8 bits x RGB). Personally, I'd much rather look at projected slides than print.
I heard a talk by a visual effects guy from ILM who explained that when they have to composite live action from film with CGI (computer-generated imagery), the first thing they have to do is compress the contrast and color range to match the range of the CGI. If there are a lot of effects in the movie, they'll end up compressing the contrast for the entire film.
Another friend of mine who recently left ILM wouldn't tell me the actual resolution they use for visual effects (trade secret), but he implied it's much lower than you'd expect.