How About a Gigapixel Digital Camera?

vcullen writes "Ever wondered where digital cameras will end up? What about a 1 Gigapixel digital camera? It would certainly beat the latest array of new digital cameras - the biggest of which only has an 8.2MP sensor! The 1 Gig Digital Camera might not quite fit in your pocket but the thought of it does make one's mind spin a little. The European Space Agency is building this massive camera (actually it's made from 170 cameras) for its Gaia space telescope, due for launch in 2010. Why? They want to map the entire universe 'down to a resolution one million times fainter than the human eye can see.'"

Comments

Pixels don't matter

[Anonymous Coward]

All that matters is the lens. By Allah, an 8.2MP camera with a quality nikon lens is better than a 1GP pixel camera with a plastic lens. When will people learn?

Re:Pixels don't matter

[Barsema]

Hm, but an 8.2MP camera with quality nikon lens is not better than a 1GP pixel camera with a quality nikon lens. zo I guess they do matter somewhat?

Re:Pixels don't matter

[Anonymous Coward]

Yeah, the four-pixel image of me is really accurate, you can nearly see the skin pores if you squint a little.

Re:Pixels don't matter

[Anonymous Coward]

Doesn't it seem strange about the 1.5 million kilometers orbit ??? Just under a million miles away... what kind of orbit it that (this is like 40 times as far away as a GEOSYNCHRONOUS satellite)... This must be a staring array, as a spinning linear array of just 10k pixels could easily equate to a Gigapixel equivalent... if the linear array also changed its pointing angle as it was spinning. Try as I might, I'm not sure how this is a great leap forward...

Re:Pixels don't matter

[ByteSlicer]

Seems to be the L1 lagrange point. A satellite here would stay in place, relative to the sun and earth. SOHO for example is also in L1 orbit.

Re:Pixels don't matter

[stevelinton]

Getting it away from Earth means that you can get long exposure more easily because Earth doesn't keep getting in the way. You can also keep things cool more easily, because you don't have a huge great IR source filling half your field of view. Finally tidal forces and variations in incoming sunlight are smaller, making fine pointing easy and saving on the batteries. The distance suggests it's going to one of the Earth-Sun Lagrange points, either between the Earth and the Sun or further away from the Sun (beyond the end of Earth's shadow cone).

As for using a filled image plane, rather than synthesising it by sweeping a sensor array across it, this is surely about how many low-brightness stars you can image and locate per hour. Given that the sources are dim (and/or you want to do spectrometry, so you are spreading the photons out according to wavelength), you don't want to waste any photons that come into the telescope if you can avoid it.

Compared to Hubble?

[FooAtWFU]

And you were complaining when they shut down Hubble! Is there anything Hubble could do that this can't?

Well, I'm sure there is. But what, praytell?

Re:Compared to Hubble?

[Detritus]

Hubble does a lot more than just imaging in visible light. How about spectroscopy?

Re:Compared to Hubble?

[Fenresulven]

Hubble can take pictues in the 2004-2010 timeframe and that is assuming this project isn't delayed or spectacularly destroyed when launched by an Ariane 5.

How about fewer, but faster

[Rufus88]

Rather than try to fit a billion pixels in a handheld camera, why not try to make sensors that operate much faster. If you could capture a hundred images at current resolutions in the same amount of time as it takes to capture a single image, you could rely on vibration-induced motion of the camera, and use motion estimation techniques to calibrate the images. Then you could use a splatting technique to sum up the images on a higher resolution grid to create an effective 100-fold resolution increase.

Of course, you wouldn't want to use a tripod with this, or perhaps you'd need a special tripod which intentionally generates random vibrational motion. Sorry if this is stupid, I'm just brainstorming here.

Re:How about fewer, but faster

[PatrickThomson]

RTFS, it's a big giant space camera. however, I read your post, and I appreciate that it's just brainstorming without thinking. I still don't care. it's made out of lots of 9 megapixel CCDs that are 4 or 5 times bigger than normal sensors.

Re:How about fewer, but faster

[Rufus88]

Dude, I *did* read the article, and I know what it was about. Nevertheless, the posted story conjured up the mental image of a gigapixel handheld digital camera by comparing it to today's state-of-the-art handhelds. That's what I was responding to.

Re:How about fewer, but faster

[PatrickThomson]

foiled again! cool idea though, isn't that what the human brain uses to increase eye resolution?

Re:How about fewer, but faster

[darkwhite]

This is definitely not the way to do it. If you need interpolation via changed lens configuration, why not shake a lens element in a controlled way? Shaking the whole camera body is just stupid, and I highly doubt the noise from that will be recoverable.

Also, you will most likely not be able to "capture a hundred images at current resolutions in the same amount of time it takes to capture a single image" in the near future. Modern sensors have massive noise as it is; they will not be able to capacitate a hundred times faster with acceptable noise without a technological breakthrough. Have you read the article? What they're doing is simply using huge lenses and putting a lot of CCD sensors (in a cryogenic, clean-room environment) next to each other. This is not achievable in regular cameras.

Re:How about fewer, but faster

[darkwhite]

This is definitely not the way to do it. If you need interpolation via changed lens configuration, why not shake a lens element in a controlled way? Shaking the whole camera body is just stupid, and I highly doubt the noise from that will be recoverable.

Also, you will most likely not be able to "capture a hundred images at current resolutions in the same amount of time it takes to capture a single image" in the near future. Modern sensors have massive noise as it is; they will not be able to capacitate a hundred times faster with acceptable noise without a technological breakthrough. Have you read the article? What they're doing is simply using huge lenses and putting a lot of CCD sensors (in a cryogenic, clean-room environment) next to each other. This is not achievable in regular cameras.

1GP, eh?

[Anonymous Coward]

How about that!

Yes But

[Itsik]

Does it have OnStar?

Simply amazing

[Anonymous Coward]

Lately, the ESA has really shown itself to have the innovation and can-do attitude that brings successes like these. What makes it all the more suprising is that they rely on a budget that is a fraction what NASA spends. Their cooporation and teamwork should be an inspiration - and a lesson - to the NASA, and US Gov't, buraucracy, who seem to thik technological and scientific advancement come by magic if you throw your money at a private company. When will they understand that blind faith in the capitalist system can only lead to their slow decline?

Put another way, it's not how much money you spend, it's how you spend it.

Been there, done that...

['t is DjiM]

Gigapixel photography using several camera's is not nearly new.

Re:Been there, done that...

[Marxist Hacker 42]

Worse than that- this uses 170 6.1 megapixal cameras, apparently. Guess now we know what a beowulf cluster of 6 MP cameras looks like.

Just so they can't point it at Earth

[dpilot]

Then the DoHS would have to kill them all.
http://yro.slashdot.org/article.pl?sid=04/09 /08/19 52241

Nitpicking

[KilobyteKnight]

It would certainly beat the latest array of new digital cameras - the biggest of which only has an 8.2MP sensor!

Not quite right

Re:Nitpicking

[FooAtWFU]

I think "the latest array" refers to the actual array which they are assembling for this project, not the cameras on the market.

Re:Nitpicking

[raider_red]

And don't forget about this guy.

Re:Nitpicking

[Too Much Noise]

First, that's not new by any definition.

Second, these Kodak cameras have larger sensors (although in this particular case larger != better quality).

Finally, there are such things as digital backs for larger format cameras. Check this one out.

Canon is nice, but it's not an end-all-be-all in photography.

Bah

[adamjaskie]

I just want a digital cam that all my manual focus Canon FD glass will mount on. Is that too much to ask?

Re:Bah

[Glytch]

One solution that comes to mind is to make some kind of adaptor to fit FD lenses on a medium format body, and use a medium format digital back. The FD mount is fairly small, and a good many medium format lens mounts are quite large, so there shouldn't be any problems machining an adaptor ring that allows focusing to infinity.

Of course, a quality digital back can be fairly pricey.

Re:Bah

[adamjaskie]

A medium format body would be too deep to focus the lenses. Even if it wasn't, the lens wouldn't cover the sensor anyway.

I've been thinking of just getting an A75 for snapshots, and keeping my AE-1, or perhaps waiting a while and saving up the $$$ to get a medium format camera and a Nikon D70.

Re:Bah

[Glytch]

Yeah, good points. The lack of full sensor coverage might not be a problem if the area covered has enough pixels, though. All depends on the specific back used. The focus problem would be a lot more difficult to overcome, though. It's a pity that Silicon Film's vaporware digital sensor never made it to market; it would be the perfect way to retrofit all kinds of manual SLRs.

And the A75 is a great little camera for the price, as long as you stay with its' ISO50 or 100 settings. The noise is very noticeable at 200 and 400.

Re:Bah

[Anonymous Coward]

I just want a digital cam that all my manual focus Canon FD glass will mount on. Is that too much to ask?

I have seen an adaptor to mount FD glass on an EF body. Should work fine with a canon digital slr. I don't remember who makes the adapter though.

Lens technology

[shaka999]

While the idea of a space telescope with such a huge sensor is pretty cool this resolution wouldn't buy much on the normal camera front.

Today many of the 8MP cameras are a joke. The lens coupled with the sensors can't come close to getting that resolution.

Increased resolution thorugh sampling...

[tod_miller]

is known as supersampling, and is used by security cameras to take several frames of footage and use this information to produce a high res version of the image.

So whilst you cannot really introduce fake information ala charlies angels and find the bad guy, if you have several frame you can clear it up substantially.

You can also undo motion blur, which is cool (I haven't the foggiest how they do it, probbaly some huffman transforms and a bit of luck!)

How About a Gigapixel Digital Camera?

Yes please.

But it isn't a gigapixel camera, that is like saying if I tie a 8 thousand 128k spectrums togather I could call it a ~~~~~gigabyte of ram.

If I did the math wrong, who cares.

What are you talking about, 8mps?

[AllMightyPaul]

You can already get 14 megapixel cameras from Kodak. And as other people have said, the pixels aren't important, it's the sensor and the lenses you use.

http://www.dpreview.com/reviews/kodakdcs14n/

Re:What are you talking about, 8mps?

[AllMightyPaul]

Ah, indeed. I see that was just announced on Tuesday, so no wonder I hadn't heard about it.

I find the idea, though, of a medium format digital camera kind of silly. But I guess it goes with the lenses for some Hasselblad cameras, as well as a few others, and as long as it doesn't mess up the field of view, then I suppose more power to 'em.

Duh

[deglr6328]

And how exactly wil ESA be getting the data from a 1G pixel survey of the universe back down here? Onboard 10Tb tape drive? Magic pixies? Sending such a vast quantity of information back down to earth with anything less than an optical frequency (read: laser) trasmitter (which dont exsist yet for spacecraft) is laughably unfeasable.

Re:Duh

[PieEye]

Maybe it can run gzip on the image successive times until it's down to a single byte, then just send that.

"Only" 8.2MP

[rawgod0122]

OK there are a few things that need to be said about that number.

First you can get cameras that have 25 MP sensors. They are called medium format. Only problem is you will be looking at tens of thousands of dollars (US) up to about $30k.

Second I have a ~6MP Nikon D70. I can print 8x10" just fine and if I had a printer large enough 11x14 with a little bit of interpolation. One just does not need that many pixels to get good prints, and even less for a computer display.

If you don't beleive me go check out
http://www.luminous-landscape.com/reviews/sho otout .shtml

and for what 3MP gets you
http://www.luminous-landscape.com/reviews/cam eras/ d30/d30_vs_film.shtml

Another nitpick

[Anonymous Coward]

The summary has a quote saying that the camera will have enough resolution to see things X times dimmer than the human eye. That's just stupid. Spatial resolution has nothing to do with light sensitivity. Light sensititivity has to do with the sensor type, the aperature size, and how much light is reflected by the lens. Resolution is independent.

Right...

[El]

So what's to stop them from pointing this fancy new camera back at the Earth? Perhaps it is not extra-solar objects that they are interested in...

How about FOUR of them?

[dbirchall]

Pan-STARRS will have a gigapixel Orthogonal Transfer CCD array on each of its four telescopes.

(The site surveys are going on right now, and I work at one of the sites being surveyed.)

If you can put one of something in orbit, you can probably put a whole lot more of something on the ground for a whole lot less money. ;)

Pixels don't matter - CCD size does

[DoctorRad]

The best quality optics in the world won't get you past the diffraction limit, so the physical size of the CCD matters too. The diffraction limit imposes the condition that cramming ever more pixels into the same area eventually becomes fruitless: there's no more information to be had in that area. You want more information? Increase that area.

That said, I have no handle on how the cost/benefit curve looks assigning funds to improving either the optics or the CCD in different proportions.

Matt...

Re:Pixels don't matter - CCD size does

[Christopher Thomas]

The best quality optics in the world won't get you past the diffraction limit, so the physical size of the CCD matters too. The diffraction limit imposes the condition that cramming ever more pixels into the same area eventually becomes fruitless: there's no more information to be had in that area. You want more information? Increase that area.

It's actually the size of the aperture, not the image plane, that imposes diffraction limits. On the image plane, the limit is imposed by the size of the beam waist you can get given the focal length of the lens and its aperture (which give the angle at which the beam converges, which gives the beam waist size for any given wavelength).

In the best possible case, you get a pixel size comparable to a wavelength of light (say one micron, for visible light). This requires a lens (or mirror!) with a diameter comparable to its focal length.

Typical pixel dimension is something like 5 microns. This is imposed by fabrication constraints (which change as process technologies get better) and by the fact that you have to have enough light falling on the pixel to produce a useful and low-noise signal (S:N improves as the square root of photon count, as long as circuit noise is low; more photons = less noise). There is a limit to how much light you can concentrate into a small region of a detector before damage occurs.

(For more information than you ever wanted to know about at least one type of image sensor, see our research group's web page.)

Re:Pixels don't matter - CCD size does

[Timothy Brownawell]

The best quality optics in the world won't get you past the diffraction limit, so the physical size of the CCD matters too. The diffraction limit imposes the condition that cramming ever more pixels into the same area eventually becomes fruitless: there's no more information to be had in that area. You want more information? Increase that area.

It's actually the size of the aperture, not the image plane, that imposes diffraction limits. On the image plane, the limit is imposed by the size of the beam waist you can get given the focal length of the lens and its aperture (which give the angle at which the beam converges, which gives the beam waist size for any given wavelength).

In the best possible case, you get a pixel size comparable to a wavelength of light (say one micron, for visible light). This requires a lens (or mirror!) with a diameter comparable to its focal length.

And since this gives a minimum useful pixel size, the only way to get more (useful) pixels is to have a bigger physical CCD. Just like was said.

Tim

Re:Pixels don't matter - CCD size does

[Christopher Thomas]

And since this gives a minimum useful pixel size, the only way to get more (useful) pixels is to have a bigger physical CCD. Just like was said.

For reasons completely other than what he said. Hence, the reply.

Onboard processing

[Anonymous Coward]

If they can afford to send an array of a billion large pixels into space, they can probably also afford to send the storage and processing power to manage it all. What they would likely do is basically just send the diffs. In other words, most bodies don't move relative to each other, so it just has to send down what changed compared to the last time that part of the sky was imaged. Of course, this is non-trivial, but I'm sure the memory and compute power is still less than what it takes to make a flight-rated 1GP camera.

aQazaQa

Re:Onboard processing

[deglr6328]

I don't think you understand what a survey is.

What are you talking about, 14mps?

[SuperBanana]

You can already get 14 megapixel cameras from Kodak. And as other people have said, the pixels aren't important, it's the sensor and the lenses you use.

What are -you- talking about? Leaf makes units that are 22mpixel, and since they go on medium-format and large format cameras which use larger and vastly more expensive/better/simpler optics, the image is often vastly superior. They also use 16 bit per channel color; the Canon 10D for example, is only 10-12(I forget which).

The Kodak 14n is an atrocious camera- the body is horrible and the sensor is so noisy Kodak had to come out with custom noise supression software that is widely regarded as the worst in the industry; up close, images look like an impressionist painting. Kodak simply smudged the image, thus negating the whole point of having a 14 megapixel sensor. The best high-pixel-count 35mm camera is the Canon 1Ds (S = studio).

The 8 megapixel consumer cameras are virtually worthless. Sensor pixel size is so small, the antialias-lens-array filter can't do a very good job and gain has to be cranked up. The Sony F828 produces images at 100 ISO that look worse than my Canon 10D at 800 ISO, and costs almost as much as the Digital Rebel, which is virtually the same as the 10D.

It's exactly what consumers deserve for being too stupid to actually look at image quality- manufactures figured out a long time ago that all the magazines were saying "more megapixels are better! Get the most you can!" That was true when cameras were 1-2mpixel. Anything over 3 is more than plenty for almost any consumer's needs. 8mpixel is gross overkill even for an 8x10 print.

Re:What are you talking about, 14mps?

[Anonymous Luddite]

It's exactly what consumers deserve for being too stupid to actually look at image quality

I'm no photographer. I did just buy a camera that said on the box "THREE MEGA PIXELS". I bought it from Walmart. I knew it wasnt't a high quality camera, (brand was Mercury I think) but figured if it was 3 MP, it must take pictures that were at least 'OK'.

WRONG.

Got it home, tried it out and got web-cam quality pictures. Total crap. returned the camera and did some research. I've bought a second, different 3 megapixel camera that is excellent, and cost about the same amount of money...

Parent is bang-on. Mod this guy way up.

Exposure times

[stevelinton]

a 1GP image at (say) 16 bits of grayscale per pixel is 2GB. Say we take these at
1000s exposure (quite short for serious astronomy, I would think), then the raw data stream is only 2MB/s. Lossless compression will reduce that pretty dramatically, although error correction and engineering data will bump it up. Plus some of the time not all of the field of view will be of interest.

Perfectly within the compass of microwave links.