Researchers Developing Single-Pixel Camera

Assassin bug writes "According to the BBC, researchers in the US are developing a single-pixel camera to capture high-quality images without the 'expense' of traditional digital photography. The idea behind such a device is that traditional digital photography is wasteful. Most of the information taken in by the camera is thrown away in the compression process. From the article: 'The digital micromirror device, as it is known, consists of a million or more tiny mirrors each the size of a bacterium. "From that mirror array, we then focus the light through a second lens on to one single photo-detector - a single pixel." As the light passes through the device, the millions of tiny mirrors are turned on and off at random in rapid succession. Complex mathematics then interprets the signals assembling a high resolution image from the thousands of sequential single-pixel snapshots. '"

Comments

Yes, it's a dupe.

[Anonymous Coward]

A Single Pixel Camera
Posted by CowboyNeal on 10-20-06 12:44 AM
from the high-tech-pointilism dept.

From the FAQ:

Sometimes I see duplicate stories on Slashdot. What's up with that?

These are just mistakes on the part of the staff. They happen. We have posted over ten thousand stories in our history. The occasional duplicate is inevitable.

If you see a duplicate, you can mail the story's author. If the story is still quiet, we may pull it down. However, once the comments are rolling in, we often leave the story up so that the discussion can continue.

Some people have suggested that there might be a software solution to this problem. If you think you've got one, visit the Slashcode site and submit a diff. As long as it isn't a performance hit, I'd consider using it. (Be aware however that the trick of searching for duplicate URLs isn't as helpful as you might think, since the same story can appear in multiple locations.)

So if you really want to complain about it, consider contributing a Slashcode patch to fix it.

Re:Yes, it's a dupe.

[alshithead]

I knew I saw this a couple of days ago on BBC. I was going to post something to the effect of "a little late on the draw there aren't we", then, I saw this is dupe. I must have missed the first post.

Re:Yes, it's a dupe.

[x2A]

Unfortunately they turned out not to be very accurate when photgraphing girls. Mirrors, as everyone knows, makes them look fat.

Re:Yes, it's a dupe.

[nomadic]

So if you really want to complain about it, consider contributing a Slashcode patch to fix it.

...Or the editors could take an extra 20 seconds to do a google search for some of the keywords. Honestly the problem isn't just that the editors don't do a minimum of checking, but that I don't quite understand what their job is if it's not to do that kind of checking. I mean, slashdot really hasn't changed much in the past 10 years or however long it's been around, what exactly are their duties so that maybe a total of a minute and a half a day can't be spared to prevent dupes?

Re:Yes, it's a dupe.

[hackstraw]

So if you really want to complain about it, consider contributing a Slashcode patch to fix it.

I've suggested multiple times for the editors just to copy-n-paste the submitted title into google like this: http://www.google.com/search?client=safari&rls=en& q=site:slashdot.org+single+pixel+camera&ie=UTF-8&o e=UTF-8

Over 99% of the time, all dupes come up as the first google link like this one. I'm a lazy MF, but I don't even see where coding this would add any value because I've seen no interest in the slashdot "editors" to care about doing any editing.

Re:Yes, it's a dupe.

[larpon]

Clearly a dupe I took a picture of the first story with my new camera . - as you can see the subject is clearly the same!

Re:Yes, it's a dupe.

[Assassin bug]

You know, FYI, I did search for this before posting. I typed "single pixel camera" in the search bar at the top of this page. You would think that such a phrase would find previous articles (and give them a high score) if the title of the article is, "single pixel camera" or even "single-pixel camera". I don't have time to contribute to patching Slashcode. But, it is kind of annoying to volunteer my time to contribute to the /.timesink and have the article posted and then have the "dupe catchers" flame up when I did what should have been sufficient to catch a previous post! Do not email me about this dupe as they suggest, I did my part to prevent it! Besides, if you consider the number of posts this dupe has received discussing the topic, maybe it needed to be brought up again.

Re:Yes, it's a dupe.

[plumby]

What I can't understand is why people complain so much about dupes. If it's a dupe you've already seen, then you can simply ignore it. If it's a dupe that you haven't already seen then you've got a second chance to catch the story.

Re:Yes, it's a dupe.

[nomadic]

I admire competence. In my work I try to do the best job that I can, I don't understand why other people don't do the same.

Reverse DLP?

[TranscendentalAnarch]

So this is kind of like a reverse DLP?

Re:Reverse DLP?

[Marxist Hacker 42]

Yep- and just as crappy as the original DLPs. DLPs work because they have *very* high processing rates- unless the subject of the picture you're taking is perfectly still, I don't see this as being a big advance in digital photography.

Re:Reverse DLP?

[Anonymous Coward]

I'd like to know how having a million "tiny mirrors" is less wasteful than having a million pixels.

Re:Reverse DLP?

[Marxist Hacker 42]

I'd like to know how having a million "tiny mirrors" is less wasteful than having a million pixels.

THAT depends on the manufacturing process- it's entirely possible that the material needed to create the mirrors is cheaper than the material needed to make a pixel sensor. However, having said that- you know how bad digital cameras are now at storing the picture once you snap the button, just based on flash memory time to store? Think a million times worse, as each pixel is scanned individually, then reassembled to store on the flash card.

Re:Reverse DLP?

[speculatrix]

basically yes. the example images are crap, they're a long way before they can produce something that'll compete with a 50$ digicam!

I would like to refer /. crowd to the fact that telescopes use mirrors and not lenses for good reasons, so perhaps a DLP mirror array with one sensor could make sense under certain circumstances... this might be an advantage is IF the image sensor had very peculiar characteristics that were very hard to reproduce in a large 2D array like a CCD or MOS sensor. Off the top of my head, that might be a very wide spectrum sensor (infrared to ultraviolet) (rather than an RGB detector) say.

Re:Reverse DLP?

[AmigaAvenger]

you do realize most telescopes have both mirrors and lenses right? a pure newtonian onto a camera/ccd is only mirrors, but every other design utilizes lenses.

Re:Reverse DLP?

[x2A]

"Think a million times worse, as each pixel is scanned individually, then reassembled to store on the flash card"

Well tha's not actually true, as the photograph isn't put together until each of the mirrors has had a turn at the sensor at least once, which means the camera must have enough fast memory to store the data from the readings and process it.

Re:Reverse DLP?

[Marxist Hacker 42]

It doesn't matter how fast the memory is- you're still only recording one pixel per whatever clock cycle the camera uses. Thus giving plenty of time for the subject you're looking at to move.

Re:Reverse DLP?

[x2A]

"I would like to refer /. crowd to the fact that telescopes use mirrors and not lenses for good reasons"

Well, they do often (or always?) still use lenses, but if you stick one mirror in there, you could half the length of the telescope (assuming you're bouncing the light the complete length of the telescope) or the thickness of the lens, and the effects are obviously greater if you curve the mirror too. Not such an issue when you have plenty of light.

Re:Reverse DLP?

[speculatrix]

this is /. so I figured brevity to keep attention span was more important than a detailed technical explanation :-)

Re:Reverse DLP?

[x2A]

How many bits at a time do you think a harddrive head can read?

Re:Reverse DLP?

[Marxist Hacker 42]

How many bits at a time do you think a harddrive head can read?

One per head, buffered. But unlike bits on a hard drive, subjects in real life MOVE. Just because you read a pixel on one side of the picture one nanosecond, doesn't mean that the next nanosecond that pixel will be the same. By using the mirrors instead of a massively parallel system, you're moving the serial from the connection to the hard drive or long-term memory storage, to actually taking the photo. Which will, at best, cause some pretty blurry photos when taking moving images. Look at the website referenced in the story- you'll see what I mean in their sample photos of even still items. The lossy compression is rotten.

Re:Reverse DLP?

[x2A]

"Look at the website referenced in the story- you'll see what I mean in their sample photos of even still items"

*pmsl* what way exactly do you think that photos of a STILL SCENE in any way reflect (hehe, reflect) image loss that WOULD be caused by taking photos of a moving scene?!!

Anyway, this isn't a simple case of turn-by-turn turning on each mirror then off again, at any one sample time multiple mirrors will be reflecting to the sensor, and for each photograph taken, each mirror will have been read from multiple times, in random (enough) order. The amount of blur you get from photographing a moving scene will be proportional to the total exposure time, as it is with any type of photography.

Re:Reverse DLP?

[vivian]

I'd like to know how having a million "tiny mirrors" is less wasteful than having a million pixels.

The real advantage this camera has is that you no longer need an array of sensors to capture a particular frequency of light. Current CCDs are really an array of pixels x 3, with one array per primary color (red,green,blue). Sure, for a regular visible light camera, a CCD is cheap to make, but with this device, if you wanted to measure some special part of the spectrum or had an unusually sensitive ( but very expensive) ccd you would only need one of them.

As you probably know, most cameras use 3 filters to capture the 3 frequencies of light close to what our eyes see - red green and blue. The part of our eyes that captures red light doesnt just capture one frequency - instead the "red filter" as it were, responds to a range of frequencies from dark red to yellowish. Likewise for green and blue, so actually we miss out on a lot of the information that is actually avaiable when looking at an object.

I imagine that with this camera, oneinteresting application woild be to directly capture full spectrum info for a scene. You could probably do something like focus the light for the single pixel sensor through a prism and create a apectrum, capturing each of the different frequency bands that way - so you would have a camera that is actually acting as a spectrograph for each individual pixel and thus capturing a lot more information than would be possible with a normal tri-filter camera.

In space applications though, cameras often capture qite different frequencies - infra red, uv, etc. With this camera, you could conceivably have an array of rather expensive specialised sensors, with each sensor being very good at capturing a particular wavelength - just as long as your mirror array can relfect all these.

Re:Reverse DLP?

[Marxist Hacker 42]

The amount of blur you get from photographing a moving scene will be proportional to the total exposure time, as it is with any type of photography.

And since serial methods of gathering data are slower than parallel methods of gathering data at the same clock speed, limiting your input sensor to a serial device will *always* create some blur. Just like in those still pictures, when they limited the input to only 4k or 8k clock cycles, the image was degraded.

Re:Reverse DLP?

[Anonymous Coward]

The main reason that telescopes use mirrors instead of lenses is because it's cheaper.

Common glass will cause different wavelengths of light to refract at slightly different angles. This means that the red light and violet light will not focus at the same spot, and you see color fringing on bright objects. It is possible to design a refractor telescope with multiple element, exotic glass lenses that works around this problem, but the exotic glass is spectacularly expensive - typically around $1000 per inch of aperture for small (less than 3" or 4") telescopes, and going up rapidly in price-per-inch for larger scopes.

Also, for larger scopes, the weight of the lenses and the length of the light path start to become problems.

Some well known companies that make apochromatic refractors (that work around the color aberations) include Teleview, Astro-Physics, and Takahashi. Compare the cost of their refractors with similar sized reflector or catadioptric scopes (the latter scopes use both mirrors and lenses.) The reason that people usually pay for the expensive refractors is they make excellent optical systems for astrophotography. They also make great planetary scopes due to better contrast, since they lack a central obstruction. But I don't personally know many people who use them strictly for visual use.

I don't see how any of this gives an advantage to the imaging detector described in the article.

Re:Reverse DLP?

[x2A]

"And since serial methods of gathering data are slower than parallel methods of gathering data at the same clock speed"

This is assuming you're moving the same amount of data. The whole point of this is that you're not, but the mirror method requires moving a lot less information from the sensor than the current method, and as the single sensor can be made a LOT better than any single sensor on a CCD array, both by improved sensitivity, and capturing light from/to a wider area, so can respond a lot quicker to the light, and so won't need the same exposure time per reading as the CCD array, making a like-for-like comparison meaningless.

"limiting your input sensor to a serial device will *always* create some blur"

On a moving scene, sure, the blur will be of size movement speed * exposure time. The lower they get, the smaller the blur, and if the scene movement speed is zero throughout the exposure, the blur size is also zero. Any non-zero blur width is purely down to implementation inperfections; inaccurate focus (standard camera you only have to get the lenses right, with this, there's millions of mirrors that all need to face/focus accurately), or diffusion, through lens clouding for example (applies to all cameras). They're technological challenges, not physical limits (that I can think of any, am open to being pointed to something I've missed. All I can think of is quantum waves theory, but this wouldn't be an issue until you tried to make the mirrors smaller/closer than the photons wavelength, as variation of where the photon lands would be the size of its wave. Microprocessors are hitting this limit (can't cut into the silicon smaller than the wavelength of light being used, but it's hard to keep increasing light frequency) - but we don't need that kind of density of resolution for a camera, and using a ccd array wouldn't change the problem here either).

Pointalism...

[__aaclcg7560]

If the one-pixel camera behaves like a traditional digital camera, I will need to take 100 pixels to get 20 decent pixels that I can use.

Re:Pointalism...

[alisson]

Film cameras are like that, too. It's really not the cameras fault, it's just tough to get a really good shot.

Re:Pointalism...

[__aaclcg7560]

With film it was more like 20 rolls to get one good shot. Very frustrating.

Re:Pointalism...

[alisson]

This is true. With digital, at least you can just delete the bad ones. But with film, you often have more choices. Of course with film, you can edit out some mistakes by hand. Of course with digital, you can edit out some mistakes with photoshop... I just said a net of 0 things.

Re:Pointalism...

[gstoddart]

If the one-pixel camera behaves like a traditional digital camera, I will need to take 100 pixels to get 20 decent pixels that I can use.

Keep your thumb out of the lens. :-P

Seriously, it's not *that* difficult to actually take decent photographs as an amateur (despite the chain of people below this thread saying otherwise). Learn a little about framing a shot, be prepared to walk a little to get in the right location, and spend a little time taking the shot. Learn your equipment -- failing that, get a fully automatic camera and spend time working on actually learning how to take good photos.

I probably take 10-20 rolls of film per year, and hundreds of digital shots. I would say that 80% or more of my shots turn out the way I wanted, and about 80% of those turned out to be pretty good shots. I'm not talking award winning photos, but useable pics. The ones I end up culling tend to be ones which turned out, but were a bit more mundane than I hoped.

Cheers

Not just for cameras

[k4_pacific]

In related news, a major roofing manufacturer has announced the "single shingle" roof. It consists of a small plate that is quickly moved about above a building during a rainstorm to block each individual raindrop. This eliminates the "complexity" of asphalt shingles.

Re:Not just for cameras

[Chandon Seldon]

That would work... if shingles were really expensive and the mechanism to move the one shingle around at the necessary speed were comparatively cheap. Oh... and you knew that you never needed to block raindrops in two places at the same time.

There are tons of ideas that work great in computerized systems that sound *really stupid* when you think of doing something that seems similar but uses other materials / technology. I mean - consider the mechanism of an ink jet printer from the perspective of a portrait artist who works with pencils...

Re:Not just for cameras

[UbuntuDupe]

Well, I was thinking, along the same lines, that this is semantics. The thousands of tiny mirrors redirecting the light are, in effect, the pixels for this camera, and it just reads them individually, aggregates them, and does some "complex" math transforms. People think of the "pixels" as the resolution, and the resolution is limited to what one mirror can put through the camera's "one pixel".

Re:Not just for cameras

[blugu64]

"There are tons of ideas that work great in computerized systems that sound *really stupid* when you think of doing something that seems similar but uses other materials / technology. I mean - consider the mechanism of an ink jet printer from the perspective of a portrait artist who works with pencils..."

Wasn't there a painted around the turn of the century that did something similar though? I can't remember the name of the artist sadly.

Re:Not just for cameras

[AK Marc]

Is http://en.wikipedia.org/wiki/Pointillism what you were thinking of? I love a good Seurat.

Re:Not just for cameras

[profplump]

http://en.wikipedia.org/wiki/Pointillism
http://en.wikipedia.org/wiki/Georges-Pierre_Seurat

Re:Not just for cameras

[Papulizer]

It is wrong to think of the mirrors as pixels. They are not read individually,
instead they are used to "integrate" the light of a given pattern.
Then these "integrals" have to be reversed to gain the image.

Re:Not just for cameras

[Agripa]

Wasn't there a painted around the turn of the century that did something similar though? I can't remember the name of the artist sadly.

This may not be what you are thinking of but I remember a series of painting that mysteriously displayed photograph like perspective that later were determined to have been made using the equivalent of a pinhole camera. The artist traced the outline of the scene while inside the darkroom on the canvas which was illuminated with the pinhole image.

On second thought, you were probably thinking of Georges Seurat's use of pointillism of which Sunday Afternoon on the Island of La Grande Jatte may be the most famous:

http://en.wikipedia.org/wiki/Sunday_Afternoon_on_t he_Island_of_La_Grande_Jatte

Being the inveterate geek, I prefer this version myself:

http://img263.imageshack.us/img263/5358/alienjatte 30dyvq4.jpg

Re:Not just for cameras

[UbuntuDupe]

It is wrong to think of the mirrors as pixels. They are not read individually

Yes, they are. Only one's light can pass through the single pixel at a time, remember?

instead they are used to "integrate" the light of a given pattern.
Then these "integrals" have to be reversed to gain the image.

I understand that. But you can already perform mathematical operations on the pixel arrays that a digital camera produces. So, the only thing this single-pixel camera does is read the mirrors separately rather than all at once, and then peform a mathematical operation on the data it gets. Nevertheless, its precision is still limited by the number of mirrors, each of which can only send it one pixel of data.

Re:Not just for cameras

[treeves]

You're probably thinking of Seurat, a pointillist, who built up his painted images from lots of little dots he made with his brush. Of course, he wasn't scanning across the canvas with three or four colored brushes dotting as he went but using some less deterministic approach.

Re:Not just for cameras

[Black Art]

Sounds like a "Breakout" product to me.

Re:Not just for cameras

[billcopc]

Well I was thinking if they're going to do this weird-ass reverse-DLP type stuff, why not take it to the next level and use a normal-sized CCD...if we could capture an 8 MP photo, twiddle some mirrors or maybe an aperture-grille-type-thing and take a dozen more 8 MP shots.. smush them together using some form of 3d interlacing type stuff and enjoy your 100+ MP shot. If you take that and scale it down to say, 50 MP you'd cut down on the interlacing artifacts and still end up with a high-rez shot for the cost of a low-rez CCD.

Hey I'm no engineer so be gentle!

Re:Not just for cameras

[phorest]

Wasn't there a painted around the turn of the century that did something similar though?

Georges Seurat is the painter I believe you are thinking of. His most famous painting (to my knowledge anyway) is A Sunday on La Grande Jatte--1884 located at The Art Institute of Chicago. There is even a Wikipedia page devoted to Seurat.

I'm sure being included in Ferris Buehler's Day Off helped to raise its profile. There is also a scale model in topiary here (very cool!)

Re:Not just for cameras

[Papulizer]

you don't read out all mirrors individually. In the link there is the following example: you have 65536 Pixels but you do only 1300 Measurements. It's not a processor that does the "mathematical operation"; it's the optics! But nevertheless you're right in one point: more Mirrors -> better Resolution.

Re:Not just for cameras

[jcochran]

Look at the 2nd link in the slashdot article.
What happens is the mirrors are "enabled" in pseudo-random patterns and the summation of the pixels seleced for each pattern is sent to the single detector. This process is repeated for each sample.
So no, the mirrors are NOT scanned one at a time.

Re:Not just for cameras

[DittoBox]

But a single, large, extremely well built and calibrated photosite is going to perform far, far better than cramming 10 million crappy ones onto chip the size of the end of your pinky.

Let's just hope that those little mirrors don't degrade image quality, lest we end up in the same boat we're in now.

Re:Not just for cameras

[Anonymous Coward]

That's not with pencils, though. That's paint. :)

Re:Not just for cameras

[ceoyoyo]

Except that a bunch of moving mirrors doesn't seem like it would be cheaper OR more reliable than what amounts to a small memory chip with part of it's casing removed.

Re:Not just for cameras

[quarkscat]

This posting reminds me about a revolutionary new file compression scheme (c. 1994) that could reduce a 2 MB MS Word
document down to 25kB, or a 2 MB TIFF file down to 40kB. There was a huge buzz in the local & national press, with
much speculation as to what the inventor's company would eventually be valued at at IPO. The big D (demonstration)
day arrived, and the software could, indeed, compress these files down to amazingly small sizes.

(Okay, okay. There IS a punch-line ... )

Reporters on the scene asked the inventor to, pretty please, decompress the sample files back to their original state.

So sorry! That compression software must be some kind of one-way hash, or (perhaps) a bit-bucket from hell. I doubt
if the public embassassment of that "inventor" went away before his 15 minutes of fame.

Re:Not just for cameras

[Tatarize]

Wow, that's a complete waste. I use a precipitation fan. Basically it is a giant fan like apparatus which blows all the water away, while it is in the air, so it can't hit my house. And my camera simply focuses the light into this cube thing, which with modern technology I use when I completely stop the light.

Re:Not just for cameras

[Chandon Seldon]

Except that a bunch of moving mirrors doesn't seem like it would be cheaper OR more reliable than what amounts to a small memory chip with part of it's casing removed.

I absolutely agree. A solid state sensor seems like it would be way cheaper than even thinking about moving parts. But... just because it seems like that would be the case doesn't mean it is - If the really clever thing here is fabricating the mirrors, they very well could be cheaper.

Re:Not just for cameras

[Dogtanian]

I remember a series of painting that mysteriously displayed photograph like perspective that later were determined to have been made using the equivalent of a pinhole camera. The artist traced the outline of the scene while inside the darkroom on the canvas which was illuminated with the pinhole image.

Possibly Vermeer?

Re:Not just for cameras

[BadMrMojo]

Seurat painted each dot in order, from upper left to lower right?

Wow, the guy was even better than I thought.

Murphy's Law

[Anonymous Coward]

Bet it'd suck to have a bad pixel with that camera, huh? :-)

Re:Murphy's Law

[imgod2u]

You joke but I think this is an actual problem. I work designing infrared cameras and the biggest thing about the testing phase of sensors is to determine the bad pixels and grade each sensor based on how many there are.

More often than not, if you have a bad pixel, you can simply turn it off or filter it out in image processing. With just one pixel, if it dies, you're SOL.

On a side note, does millions of micro-mirrors sound more difficult and prone to failure than a million photodetectors to anyone else?

Re:Murphy's Law

[scdeimos]

It'll be a darn side worse with this.

Consider not only the possibility of the single pixel image sensor failing, but the possibility of one or more of the MEMS mirrors getting stuck in the "on" position. A single mirror would cause the entire image to brighten (by how much depends on what it's looking at), multiple pixels could actually saturate the image. Mirrors stuck in the "off" position wouldn't pose such a problem - they'd present a "black" pixel in the output image but wouldn't affect anything else.

Re:Murphy's Law

[plover]

Here's my previous post on the subject:

Micromirror arrays have been commercially available for ten years now, and had been in design for at least ten years prior to that. They're used in DLP projectors and projection TVs. You can go buy one at Best Buy if you'd like.

The durability of a micromirror array is actually very high. It's counterintuitive, but not hard to understand. The reason is the mirrors are so tiny. They have very little mass which means they transfer very little stress to their mechanical structure, even under large G force loading.

Think about the normal operating conditions of a micromirror in a DLP TV -- each of those mirrors is designed to flap at 100 kHz. They're already subject to extreme G forces in their everyday operations. Bouncing a chip off the ground is not much force compared to actually using it.

A good question would be the efficiency of light transmission. There's a clear shield mounted over the mirror array, which will attenuate the light both on the way in and on the way out. And the mirrors themselves can not be 100% efficient reflectors. But I suppose with a single pixel detector, you can invest more in making it very sensitive to low light conditions.

Finally!

[Anonymous Coward]

Now we can get pr0n at the level of quality in Duke Nukem! One fleshy-pink-colored pixel is enough to get most me off...

Re:Finally!

[voice_of_all_reason]

Dude... aliens in disguise. You could even see the green tentacles sticking out.

Re:Finally!

[Anonymous Coward]

One fleshy-pink-colored pixel is enough to get most me (sic) off...

Best typo ever.

Re:Finally!

[Anonymous Coward]

Haha, thanks! I tried to correct myself right after that but slashdot has their time limits...

Re:Finally!

[Anonymous Coward]

Cut him some slack, he was probably typing one handed while masturbating over the dot on an i.

Comment removed

[account_deleted]

Comment removed based on user account deletion

RAW format anyone?

[Anonymous Coward]

> Most of the information taken in by the camera is thrown away in the compression process.

Doesn't the RAW format take care of this?

Re:RAW format anyone?

[John Meacham]

The problem is not getting at that extra information, like you say, we can already do that with RAW. the problem is that a lot of resources (such as CCD area) go into capturing this extra information which is then simply discarded. By taking a random sampling of pixels, one gets exactly as much information is needed to construct the compressed version of the image without waste. plus, with only a single CCD, you can make it incredibly sensitive, to the point where it can count single photons. Heck, you could probably have some fun with wavelengths. different wavelengths get diffracted slightly differently, if you could take advantage of that to redirect photons of different wavelengths at the sensor. you could have a camera that takes _full spectrum_ pictures. not just at the single pretty but not very informative red green and blue lines. (tetrachromats rejoice!). Full spectrum sampling in a small package would be really cool, I mean, that is tricorder technology. This is very neat research.

Re:RAW format anyone?

[Pieroxy]

There is always a catch, however. Let's take an example of a 1MP camera, taking a picture at 1/100th of a second. Each CCD can acquire light for a full 1/100th of a second. But each one is small and as such, not very sensitive.

Let's say this new 1 pixel camera is set-up to take a picture of 1MP at 1/100th of a second. Each one of the 1M mirrors will reflect its light on the CCD for ... (1/100)/1000000 th of a second, because only one pixel (of the final image) can be recorded at a time. So yes, the new sensor will be more sensitive. And it better be ! 1 000 000 times to be correct (for 1MP pictures.)

Re:RAW format anyone?

[Goaway]

Incorrect. Even the summary implies that it lets a lot of different pixel shine on the sensor at any moment, and untangles it with maths afterwards. It still needs to be faster, but there's also just one, and it can be big.

Re:RAW format anyone?

[dangitman]

the problem is that a lot of resources (such as CCD area) go into capturing this extra information which is then simply discarded.

What do you mean discarded? I use every pixel of my digital SLR in RAW mode, and I often wish there were more pixels. A lot more. So, where am I discarding these pixels?

Re:RAW format anyone?

[Thuktun]

Heck, you could probably have some fun with wavelengths. different wavelengths get diffracted slightly differently, if you could take advantage of that to redirect photons of different wavelengths at the sensor. you could have a camera that takes _full spectrum_ pictures. The limiting factor there would be finding a material that allowed you to reflect or diffract photons all the way from infrared light to x-rays. Since we're talking about a single-pixel sensor, constructing separate sensors for the various areas of the EM spectrum would probably be more feasible.

Re:RAW format anyone?

[timeOday]

go into capturing this extra information which is then simply discarded.

The information is not simply discarded, even when using lossy compression. Rather, it is selectively discarded. If you can closely approximate a whole series of data with a more compact formula, you do it. But to know whether that's the case, you first have to gather and examine the information! Otherwise you cannot know what you might be missing.

Re:RAW format anyone?

[ceoyoyo]

In order to capture as much information as a one megapixel image, this camera will have to take one million different measurements. So the poster you replied to is correct in principle. They save a bit because they can capture the information in a compressed form (artifacts and all), but they still have to make a LOT of measurements, very quickly. I expect the noise characteristics are horrendous too -- if you have a bit of movement while you're making all those measurements then you're going to end up with a very interesting looking picture once you're done doing all that processing.

Re:RAW format anyone?

[ceoyoyo]

They're talking about the people who use their ten megapixel point-and-shoots to take out of focus pictures in the dark and then post them as 500 pixel wide pictures on MySpace.

Re:RAW format anyone?

[dangitman]

So, why generalize? That does not make digital cameras inefficient. It makes the users of digital cameras inefficient. They shouldn't claim to be making more efficient cameras, they should be claiming to make cheaper crappy cameras for the Myspace folk. But not every camera user uses their camera like this.

Re:RAW format anyone?

[ceoyoyo]

I'm right with you. I shoot RAW exclusively.

MySpace folk don't like to think they're buying crappy cameras though, so you have to make them think it's the latest and greatest thing and then they can laugh at all the SLR owners who paid ten times as much for their cameras, have fewer megapixels and have to look through those old fashioned little windows to take a picture!

Re:RAW format anyone?

[Anonymous Coward]

can you clarify that statement because all I'm seeing is "As the light passes through the device, the millions of tiny mirrors are turned on and off at random in rapid succession. Complex mathematics then interprets the signals assembling a high resolution image from the thousands of sequential single-pixel snapshots."
note key words/phrases
"succession" viz a number of people or things sharing a specified characteristic and _following one after the other_
"sequential"
the example above introduces another issue that would occur taking pictures of moving objects - distortion introduced by the sampling - even if random - would introduce hazing - though probably not a dramatic but effectively as severe as scanning back cameras where movment introduces surreal vertical or horizontal smearing (depending upon the orientation of the scanning pass).
anyway while I think the "single pixel" camera would be of limited, but useful, application the technology would be applicable in standard digital photographic processes in allowing greater total effective megapixel count on the same size sensor, and/or allowing larger sensors (which are more sensitive, less susceptible to noise and interference from adjacent sensor pixels) without introducing (or at least mitigating) many of the issues that arise from cramming more and more pixel elements in to a fixed size sensor.

Re:RAW format anyone?

[Anonymous Coward]

... hey wait I have a myspack...errr myspace... page and I shoot only RAW on a DSLR... I didn't know that these were mutually exclusive.
I feel like such a freak now..... no... wait... I've always been a freak :P

Re:RAW format anyone?

[ceoyoyo]

Freak. Pick a side!

Like I should talk.

Re:RAW format anyone?

[iangoldby]

can you clarify that statement because all I'm seeing is "As the light passes through...

This camera was described on the recent BBC Digital Planet programme on the BBC World Service. They explained in the program that (a random) half of the mirrors are 'on' at any one time.

Having said that, I agree that the sensor still needs very good sensitivity and noise properties, because to disentangle the multiple images, some of the maths will involve subtracting one image from another, and that will amplify the noise. I wouldn't be surprised if the net result was the same as if only one mirror was on at a time (leaving aside of course the main point of the camera which is to only record as much information as will go into the final compressed image).

The question is, can a single large detector be a million times better in terms of sensitivity/noise than a million small detectors? Maybe...

Re:RAW format anyone?

[zacronos]

In order to capture as much information as a one megapixel image, this camera will have to take one million different measurements.

You mean "In order to capture as much information as a one megapixel image with no compression in the final image". In contrast, what this camera is doing is essentially only capturing as much information as is necessary to make the final, compressed image. It may be one megapixel in the sense of having ~1M pixels in the final image, but if it's been JPEG compressed to 80% then it doesn't need ~1M measurements.

Re:RAW format anyone?

[f00Dave]

Who said only one "virtual pixel" was going to be reflected at a time? A quick skim of TFA reveals that, for any sample, some set of mirrored beams will be engaged, not just one. We're really obsessed with our "one beam to rule them all" thing, culturally, aren't we? ;-)

Re:RAW format anyone?

[ceoyoyo]

Yes. Did you read the next sentence of my post?

"They save a bit because they can capture the information in a compressed form (artifacts and all), but they still have to make a LOT of measurements, very quickly."

Also note that when doing regular JPEG compression the image you're working on doesn't change. When doing direct compression the image you're compressing WILL change. That's going to have some serious consequences.

Also, JPEG doesn't always compress by the same factor. A complex image doesn't compress as well. How do they handle that? The difference can be huge... if I ask this camera for a 1/100 of a second exposure and the scene is more complex that it thinks does it have to keep measuring for 1/25th of a second?

Replacement policy?

[Anonymous Coward]

I'd find out what the dead pixel return policy is before buying one of these.

complex mathematics?

[superwiz]

Surely, you mean "complicated". Mathematics already has a use for the word "complex".

Re:complex mathematics?

[Jerf]

It's a little clue for those "in the know" that the described benefits are entirely imaginary.

Re:complex mathematics?

[rherbert]

Perhaps they're implying that the mirrors move in Hilbert space, which would be quite an accomplishment.

Re:complex mathematics?

[Anonymous Coward]

That's funny, I thought we were talking about cameras, not maths.

Language has this funny capability of ascribing different meanings to the same work under different contexts.
In this case, the mathematical one is the aberration.

Re:complex mathematics?

[CrabbMan]

Mathematics already has a use for the word "complex".

Actually, mathematics has many uses for the word "complex," not all of them even having to do with imaginary numbers. Don't even get me started on "normal." The trick is, you have to interpret the meaning based on context (which is really the case for any word with multiple meanings, mathematical or not), and in this case I think the context makes it obvious that "complex" should be interpreted as "complicated."
Sorry for the waste of space here; I suppose my only point is that it ain't necessary to correct what don't need correcting (and I'm probably at least half-way to being a hypocrite here).

Re:complex mathematics?

[x2A]

Yes, but the word was not used within the mathmatical context.

Re:complex mathematics?

[superwiz]

You are right... mostly. If someone said "complex analysis", they would also be talking about math rather than talking math. The think of it is, when you talk about interpolating and extrapolating partial results that involve wave-form, you are very likely to use complex analysis. If for no other reason that complex numbers describe waves better (I really mean easier, less complicated, to avoid the pun) than the real ones do. So it is entirely possible that the acurate description of the process used to store pictures was to say that some complex (as in "involving sqrt(-1)") mathematics was used. Complex numbers are routinely used when describing waves because a complex number can be used to incororate both amplitude and phase of a wave.

Re:complex mathematics?

[superwiz]

Well, I could hope that you get modded as flaimbait, but I think it's better to just explain why you are wrong. Complex numbers are routinely used to talk about waves. Which could be relevant when talking about light. If you are not familiar with a concept than don't jump the nerds who talk about at a place which brands itself "news for nerds". We are here because we like ponder the details of how things work. If you don't, let us be or remove yourself.

Re:complex mathematics?

[WolfWithoutAClause]

That's not entirely true, they're quite complex; but I won't explain the benefits here. They have a real component.

Throwing away data?

[kerohazel]

Well, there's no reason a digital camera *has* to throw away any data at all. It's likely the case that all digital cameras do perform on-the-fly JPEG compression, but it's not a limitation of the hardware, so why bother reinventing the wheel if you really care about losing data that much? Just make a digital camera that saves pictures as some lossless format.

And at any rate, how are the single-pixel cameras throwing away any *less* data than their plain digicam counterparts? Doesn't it all just depend on the encryption method used?

Re:Throwing away data?

[Anonymous Coward]

When researchers base an attack on how a competing method works, it's usually because they lose if they make a direct comparison of the results.

Re:Throwing away data?

[x2A]

Tiny pixels have a chance of being triggered by a single stray photon (or something), causing speckles on the image. You can figure out the anomalies by taking multiple readings from each pixel, and discarding any that are very different from the rest, but that can be slow. Or, you can discard based on pixels very different from their surrounding pixels, and replace with the average of surrounding pixels, and then you start to lose picture quality.

Using lenses, each mirror can capture light from a larger area that it takes up, and reflect it to a larger sensor, which means it's not going to be as affected by single stray photons.

(I pretty much made all that up on the spot, but it sounds right)

Re:Throwing away data?

[nospam007]

Well, there's no reason a digital camera *has* to throw away any data at all. It's likely the case that all digital cameras do perform on-the-fly JPEG compression, but it's not a limitation of the hardware, so why bother reinventing the wheel if you really care about losing data that much? Just make a digital camera that saves pictures as some lossless format.
--
It is able to (re)construct the image with a reduced number of _random_ pixels taken, so instead of 'throwing away' unneeded info, it just doesn't need to _get_ that info in the first place to be able to display the image.
So it should be able to work faster and not slower than the traditional way, just as inkjet printers are faster than plotters.
(once it has the same development stage that is)

Re:Throwing away data?

[ceoyoyo]

Using a lens your conventional CCD or CMOS element registers light from more than it's own area as well (that's one of the purposes of a lens). The only difference in this case is that you've stuck mirrors in the way, which will absorb some of the light. This sensor won't be any less susceptible to noise. Remember, because you have to change the pattern a few million times during the exposure that single sensor has to make several million individual measurements instead of being able to sit and collect light over the entire exposure time. Except in this case the noise will get transformed by the "complex math", and consequently amplified and cause strange artifacts.

Need help making sense of this...

[posterlogo]

FTA: Compressive Sensing is an emerging field based on the revelation that a small group of non-adaptive linear projections of a compressible signal contains enough information for reconstruction and processing. We have developed algorithms and hardware to support a new theory of Compressive Imaging. Our approach is based on a new digital image/video camera that directly acquires random projections of the signal without first collecting the pixels/voxels. Our camera architecture employs a digital micromirror array to perform optical calculations of linear projections of an image onto pseudorandom binary patterns. Its hallmarks include the ability to obtain an image with a single detection element while measuring the image/video fewer times than the number of pixels --- this can significantly reduce the computation required for video acquisition/encoding. Because our system relies on a single photon detector, it can also be adapted to image at wavelengths that are currently impossible with conventional CCD and CMOS imagers.

Is this like interpolation? Stacking? Averaging? Can't figure out WTF they're talking about, but it sounds like you have a single pixel camera, but that it must acquire many readings (sequentially). The number of readings required will still be less than the final pixel could your image will end up at. Hmm... seems like it would have to make up some of the information that it did not actually acquire?

Re:Need help making sense of this...

[kimgkimg]

This could definitely increase the light sensitivity of a camera since you're not tied to trying to cram millions of photosites into a given chip area. Does anyone remember the linear scanning photography that people use to hack up by mating a linear scanning mechanism from a flatbed/handheld scanner onto the film plane of a camera? Similar idea except they are using an array of mirrors instead of the linear motion of a CCD array. I wonder what the differential cost of mirrorchip + one-pixel sensor vs. multi-megapixel CCD chip is going to be.

Re:Need help making sense of this...

[LiquidCoooled]

They have a single sensor and mega-mirrors which all flutter about.
At any one point in time, only a minimal number of mirrors will actually send a reflection onto the sensor.

they use a mathematical algorithm to process these and reconstruct the multi mega pixel image back from the positions on the multi mega mirror array.

Its another case of smoke and mirrors.

Re:Need help making sense of this...

[Anonymous Coward]

They use multiple measurements. But the key is, that you don't need to reflect light from a single mirror to the sensor, but can reflect from multple at the same time, in effect observing the sum of light reflected from all mirrors.
The following key idea is that they use a new mathematical method that states that under some circumstances you can reconstruct a compressed signal from observing inner-products of the (uncompresed) singal and random signals.

Basicall, if you can compress your image (using a specific kind of compression, i.e., non-linear approximation), of M pixels to N pixels (N M), then you can in theory reconstruct your compressed signal from cN observations (with c some constant).

Re:Need help making sense of this...

[0xABADC0DA]

It's more like linear algebra... you have a series of equations based on the random-ish mirror patterns, so for a simple 2x2 grid of pixels (mirrors):

1*x0y0 + 0*x1y0 + 1*x0y1 + 0*x1y1 = sample1
1*x0y0 + 1*x1y0 + 0*x0y1 + 1*x1y1 = sample2
0*x0y0 + 1*x1y0 + 0*x0y1 + 0*x1y1 = sample3 ...

Then you basically solve it for the pixels. So think interpolating the entire image at once as a single value.

Re:Need help making sense of this...

[x2A]

"then you can in theory reconstruct your compressed signal from cN observations"

THE SPEED OF LIGHT TIMES N?!! That'll take forever!

"(with c some constant)"

Oh...

We'll See...Betamax anyone?

[MuChild]

1) Create a million bacteria-sized mirrors. 2) ???? 3) Profit!

Single pixel reflector telescope

[heroine]

Always thought the single pixel idea would be more practical in a reflector telescope. Such a telescope could have a much higher dynamic range than any other telescope due to the extra money available for the pixel. The telescope would use the Earth's rotation to scan one axis and servos to scan the other axis.

Re:Single pixel reflector telescope

[Cecil]

That would require *extremely long* exposures. You can't really up the sensitivity of the detector no matter how much money you throw at it. Good CCDs are already at the level of being able to record single photons, and it still takes hours or even days for some exposures. So what's a several hour exposure multiplied by a million pixels? Ouch is what it is.

Re:Single pixel reflector telescope

[Hijacked Public]

You can't really up the sensitivity of the detector, but you can change what you do with the signal it sends once you get it.

With current pixels densities (say about 16MP on a full frame 35mm sensor) the signal can be boosted to give the ISO equivalent of 3200 speed film before the noise becomes objectionable. Noise is a problem in part because signals from adjacent pixels affect one another, which is magnified when the signal is boosted to ISO 3200.

I would suppose that a single pixel would be less prone to noise when boosted.

Hot or stuck pixel?

[jo7hs2]

Oh great, now I'll end up with a camera with a stuck or hot pixel and be totally screwed. Thanks, progress.

Re:Hot or stuck pixel?

[mobby_6kl]

Well, you could always use the dark frame subtraction method to fix the problem.

Re:Hot or stuck pixel?

[cfulmer]

Actually, I don't think you could. If you had a mirror that got stuck into the 'on' position (i.e. it's pointing at the single sensor), it would partially blind the sensor whenever any other mirror was also pointing at the sensor. If that one mirror happened to be seeing pink, the entire photo would have a pinkish hue. If it happened to be seeing white, the picture would be washed out. If it happened to be seeing pitch black, well, then you're in business.

Re:Hot or stuck pixel?

[x2A]

Nah, you just turn off all the mirrors (or try to), take the reading, and subtract that value from all future readings.

Re:Hot or stuck pixel?

[cfulmer]

But, that's the problem -- that "value" will change based on what it's projecting. It's not as if the sensor is stuck at sending "0x5c" or something as it is in a normal digital camera -- it's stuck at adding mirror #41956's current output to the output of all the rest. And, that output can change. It could be that the correct answer is to sample mirror #41956 and then subtract that from the results of all the other pixels (if that even makes sense), but that ends up severely diminishing the range of the others: if the mirror happens to be sending white light (say 0xffffff), subtracting that from the results for the other mirrors is going to give you a black image, except for one pixel.

Re:Hot or stuck pixel?

[x2A]

"if the mirror happens to be sending white light (say 0xffffff), subtracting that from the results for the other mirrors is going to give you a black image, except for one pixel"

No cuz each reading is going to be of a lot of pixels on at a time, so the sensor has to be capable of the maximum amount of light you'd get from *all* of those pixels being brightest, which lets face it, ain't gonna happen much.

And each mirror pixel can pick up light from a wider area compared to using a CCD, because with CCD the focal point has to be where the sensor is, ie the pixel is, whereas with mirrors, it's not.

Re:Hot or stuck pixel?

[x2A]

"And each mirror pixel can pick up light from a wider area compared to using a CCD, because with CCD the focal point has to be where the sensor is, ie the pixel is, whereas with mirrors, it's not"

Sorry ignore that bit, jumped between writing replies and carried on typing in the wrong window *lol* wanted to quickly finish it as a friend arrived

It's the Mirrors

[Ponga]

That cute little girl has been saying it all along!

Ugh, I need to stop watching so much TV...

Re:It's the Mirrors

[Joe Snipe]

Juh? What does that mean?

it's called "Compressed Sensing"

[toby]

And this story hit the UK Guardian on 9 Nov 2006. (via CS maven my slice of pizza.)

I used it for my holiday snaps

[Timesprout]

This is me skydiving
.

This is me swimming with dolphins
.

This is me at the grand canyon
.

Re:I used it for my holiday snaps

[Anonymous Coward]

This is me having sex with my wife (NSFW)

.

Re:I used it for my holiday snaps

[MindStalker]

Dude, you should link stuff like that and no post directly.. I really didn't want to see your ugly whale of a wife..

Re:I used it for my holiday snaps

[Anonymous Coward]

Your nipple's showing.

Re:I used it for my holiday snaps

[Asgerix]

That's the good thing about dupe posts; you get to improve on your jokes!

http://science.slashdot.org/comments.pl?sid=201687 &cid=16513113

Re:I used it for my holiday snaps

[LordNimon]

This would be great for pictures of Madonna.

Re:I used it for my holiday snaps

[Anonymous Coward]

That's the good thing about dupe posts; you get to improve on your jokes!
http://science.slashdot.org/comments.pl?sid=201687 &cid=16513113

1) A dupe article
2) A dupe post
3) What's next? A dupe profit?

Re:I used it for my holiday snaps

[Krommenaas]

bah, black&white photography is so arty-farty.

Re:I used it for my holiday snaps

[Scoutn]

I think you should take issue with your travel agent; your adventures all look the same to me. (Even those hidden pictures you call dots above the "i"'s.)

Re:I used it for my holiday snaps

[x2A]

WOW... that post you linked to was dated Oct 20 (or 16 in dec) ...

don't get out much huh?

Re:I used it for my holiday snaps

[picob]

Mod parent dupe up.

Re:I used it for my holiday snaps

[Anonymous Coward]

Panoramas are more interesting: ........

Single-pixel DLP-type camera is cool because...

[inviolet]

...with only a single CCD pixel, they can spend all their resources making it exquisitely sensitive, so as to outperform normal array CCDs.

Of course, they'd have to do that anyway, because to get a decent shutter speed they're already going to have to 'scan' the viewed area extremely quickly. It's the old tradeoff of serial versus parallel processing.

Re:Single-pixel DLP-type camera is cool because...

[radarsat1]

I'm gonna sit and wait until they perfect this, and just before it gets popular (because its so cheap) I'm going to patent the 2 pixel camera with twice the resolution for only a tiny higher cost, and beat them at their own game!

Mouahahahah...

Re:Single-pixel DLP-type camera is cool because...

[Anonymous Coward]

Can you imagine a Beowulf cluster of single-pixel ...... oh, wait......

Re:Single-pixel DLP-type camera is cool because...

[Jeremi]

I'm gonna sit and wait until they perfect this, and just before it gets popular (because its so cheap) I'm going to patent the 2 pixel camera with twice the resolution for only a tiny higher cost, and beat them at their own game!

You joke, but it might be a good idea... there's no need to stick to either the one-CCD-pixel-per-camera or the one-CCD-pixel-per-image-pixel extremes. Perhaps there is a happy medium somewhere, like having 256 scanning-CCD-pixels operating in parallel to build up a (simulated) 16-megapixel image?

Re:Single-pixel DLP-type camera is cool because...

[Anonymous Coward]

>...with only a single CCD pixel, they can spend all their resources making
>it exquisitely sensitive, so as to outperform normal array CCDs.

Ever think about the thermal noise limitations? .... Didn't think so.

Pray tell how are you going to get color? The mechanical spinning wheel as used in DLP projectors? The ones that give you great contrast, but horrible color saturation?

Re:Single-pixel DLP-type camera is cool because...

[x2A]

The resolution would be defined by the number of mirrors so HAHAHAH take that :-p

Ah, more moving parts. THAT's helpful.

[ScentCone]

Is it just me, or does the concept seem inherently more complex and fragile than a multi-pixel sensor with light cast on it?

And how can this possibly deal with the equivalent of a range of shutter speeds in front of a standard sensor? Perhaps it's a matter of how many times the pixel is exposed to the same part of the lens' projection in repeated scans... but that just seems clunky, and that much harder/slower to re-assemble into a stored image.

And it doesn't stop the megapixel chest thumping - it just starts up megamirror arguments, instead.

Re:Ah, more moving parts. THAT's helpful.

[Jeff DeMaagd]

Micromirrors are actually very reliable and even exceed the lifetime of a typical LED now, of hundreds of thousands of hours of constant flexing. It turns out that nano-scale objects have different properties. A piece of metal on the nanoscale is likely to be a single crystal and that usually eliminates the fatigue issue. I think this has more uses in the sciences though.

Re:Ah, more moving parts. THAT's helpful.

[jandrese]

It still sounds to me like a camera with literally millions of moving parts though. Also, the sheer physics of getting the photons to the sensor (especially in low light conditions) when each pixel only has the most absolutely tiny slice of time to collect light seems rather difficult to surmount. In low light conditions it's easy to see where this could amount to only a handful of photons per pixel. It's hard to see how you wouldn't get lots of noise in low light conditions just due to the magnified effects of random variation in the number of photons hitting a lens per second.

Re:Ah, more moving parts. THAT's helpful.

[voice_of_all_reason]

Stating the researchers are talented is just as much an assumption as suggesting the GP is from the peanut gallery. Maybe they deserve turds.

Re:Ah, more moving parts. THAT's helpful.

[Goaway]

Just when you thought Slashdot couldn't get any more idiotically arrogant...

Re:Ah, more moving parts. THAT's helpful.

[pla]

And how can this possibly deal with the equivalent of a range of shutter speeds in front of a standard sensor?

Not to mention, you get the previously-not-an-issue joy of temporal aliasing.

And you thought the flicker of fluorescents annoyed you now? Wait until any exposure longer than 1/120th of a second includes both the "lit" and "unlit" version in one picture. Good luck figuring out the meaning of white and black levels on that monstrosity...



And it doesn't stop the megapixel chest thumping - it just starts up megamirror arguments, instead.

Well, realistically, you wouldn't need millions of mirrors, just one, which can tilt in 2d. It would thus "scan" the target just like a CRT scans the screen - Except, instead of hitting distinct pixels on its path, you'd have a continuous function.

In theory, you could literally generate an infinite-megapixel image from that. In practice, you'd just replace "megapixels" with "megahertz" - The effective resolution would depend at the frequency with which you could sample the actual sensor. You'd need a sampling rate of 1GHz to get an 8MP image at an exposure of 1/125s (incidentally, that applies whether you use one 2d mirror or an array of 8 million binary mirrors).



I do seriously have to wonder, though - Considering how poorly current digital cameras perform in low-light conditions, wouldn't this massively worsen the problem? Even taking a picture of the Sun itself, does a 1/(10^9) exposure really give a meaningful value for a single pixel?

Re:Ah, more moving parts. THAT's helpful.

[x2A]

"And how can this possibly deal with the equivalent of a range of shutter speeds in front of a standard sensor?"

Shutter has to move a greater distance, these only gotta move TINY distances, and being so small, a tiny electromagnetic (or something) field can completely open or close it in a very small amount of time. Compare that with having to move an entire shutter. So, you've got smaller things to move, and smaller distances to move them by. Sounds line win * win (win^2) to me.

Re:Ah, more moving parts. THAT's helpful.

[x2A]

Well the mirrors actually more like flicker throughout the the period of exposure, so have multiple chances to detect light. Also, with the focal point on the CCD on a standard camera, each sensor can only collect light from the area it covers, whereas a mirror will be collecting light from a larger area than it's size, and focusing it all onto the sensor, more than making up for the fact it gets shorter total amount of exposure time.

Re:Ah, more moving parts. THAT's helpful.

[GigsVT]

Hah!

I run NTSC into the open source motion program for home security. When a wasp checks out the camera (often), their wings are beating so fast that half the scan lines have the wings up and half have it down.

So with interlaced signals we already do get some temporal aliasing. :P

Re:Ah, more moving parts. THAT's helpful.

[uptownguy]

Micromirrors are actually very reliable and even exceed the lifetime of a typical LED now, of hundreds of thousands of hours of constant flexing. It turns out that nano-scale objects have different properties. A piece of metal on the nanoscale is likely to be a single crystal and that usually eliminates the fatigue issue. I think this has more uses in the sciences though.

Exactly. Think Hubble-sized.

Re:Ah, more moving parts. THAT's helpful.

[Overzeetop]

Yes, and since TI owns much of the patent base for micromirror stuff, we can half half the effective pixels for the same price, and glacially slow advancement, just like we get from them for projectors. WooHoo!

I have to say it

[JohnnyGTO]

oops, crash sevem million years bad luck !?!

Urgh!

[HerrEkberg]

Look how many MegaMirrors my new camera has!

Excuse me

[markov_chain]

Please don't move until I sequentially activate a few hundred thousand micromirrors!

'nuff said.

Dupe

[rumith]

From the mysterious past: http://science.slashdot.org/science/06/10/19/22552 39.shtml.

Sigma-Delta Modulation

[sitturat]

It sounds very much like Sigma-Delta Modulation (http://en.wikipedia.org/wiki/Sigma-delta_modulati on). Lots of samples in time, fewer bits.

Contradicts itself.

[Xoltri]

The article says that this new camera will have do do "Complex mathematics to interpret the signals" but at the same time will "do away with the need to process and compress each image". So which is it? I just don't see how this will save anything if you have 1 pixel doing something 5 million times or 5 million pixels doing something one time.

Re:Contradicts itself.

[blackholepcs]

The article says that this new camera will have do do "Complex mathematics to interpret the signals"

This camera will take awesome pictures AND it comes with doodoo? SWEET!

Re:Contradicts itself.

[Antitorgo]

Sort of -- it does the "compression" pre-capture by the CCD. They have random patterns in the mirrors hitting the CCD and use a DSP to reconstruct the image based on what the random patterns ended up being. It is kind of an interesting concept. The other benefit is that more money can be put into a higher quality CCD (such as one that senses in UV or IR).

I'm kind of interested in why the patterns have to be done randomly (or pseudo-randomly). That mean that two shots using the exact same settings could produce vastly different results. I imagine that with long enough "exposure" (really collection) time that the imagines would end up being basically the same. From what I can gather, the # of samples taken would be determined by the exposure time and that the DSP is adjusting the brightness levels post-processing. So longer exposures would give you a clearer picture versus what you'd traditionally see as over/underexposed. So a shorter exposure time would give you a blurrier picture (as seen in the second link).

Re:Contradicts itself.

[dangitman]

The other benefit is that more money can be put into a higher quality CCD (such as one that senses in UV or IR).

Say what? Don't CCDs inherently respond to UV and IR? The problem is filtering out the IR and UV so it doesn't affect the visible spectrum and therefore your normal photos. This is why cameras have IR and UV filters over the sensor. To modify a standard camera to shoot IR for example, you simply remove the IR filter. And add an IR-pass filter if you want pure IR, rather than visible+IR - but that's not necessary.

Re:Contradicts itself.

[Antitorgo]

Well, I was using the broader wavelength capacity as an example, not necessarily as a valid argument. My point is that you could substitute a higher quality CCD with a higher dynamic range for example (registering more levels of light) or have a 3CCD setup for very little cost...

How about that you could essentially make any CCD into a frame transfer CCD by focusing light onto different parts of a CCD?

Re:Contradicts itself.

[x2A]

No dummy, if you bothered to take into account the context, 'do do' is obviously being used in its adjective form, not noun. All cameras do complex maths, this cameras complex maths just happen to be poo.

Re:Contradicts itself.

[x2A]

"I'm kind of interested in why the patterns have to be done randomly (or pseudo-randomly)"

Because you want each reading to be as different to all other readings taken for the shot as possible, so the best "random" generators (one's that produce the most convincing white-noise) are statistically gonna be about the best way to do this. If you did it pattern based, you'd possibly have to put the extra work into compensating for it so you don't get image artifacts (although not really sure about that).

"That mean that two shots using the exact same settings could produce vastly different results"

Each mirror is flickered enough times in different combinations to cancel that out - would have to be, if you can get different results, then it can't be accurate enough to useful.

Re:Contradicts itself.

[Anonymous Coward]

Actualy this is a very comon mis-typing for people having dyslexia or similar deficiencies.

Re:Contradicts itself.

[Kuciwalker]

It could use complex mathematics to do away with the need for (much) processing.

Re:Contradicts itself.

[Anonymous Coward]

The article says that this new camera will have do do "Complex mathematics to interpret the signals" but at the same time will "do away with the need to process and compress each image". So which is it? I just don't see how this will save anything if you have 1 pixel doing something 5 million times or 5 million pixels doing something one time.


The concept does seem incomplete, somehow. Though there's lots of mirrors (a million of them!) but there's no smoke. Where does the magic smoke come into play and where does it exit?

Re:Contradicts itself.

[Anonymous Coward]

They don't take 5x5 million samples but only - say - 100000 pseudorandom samples. If you want to reconstruct the image you have to optimize a certain functional which is costly but can be done offline. Because of the redundancy of a natural image these samples contain all data necessary to reconstruct the full image. You can view this as a lossy compression algorithm which offloads all computation to the decoder. The encoder is so simple that you can combine it with the measurement process (sampling at pseudorandom locations). It is known as "compressed sensing" in the mathematical literature. We are working on similar techniques in MRI to reduce scan time.

-- Martin Uecker

Still patented too

[goombah99]

The coherent detection version of this was patented 11 years ago.
Apparatus and method for heterodyne-generated, two-dimensional detector array using a single detector

Re:Still patented too

[Rei]

At first, I thought this was going to be similar to the method of generating hires images from a small number of sensors utilized by jumping spiders. Basically, they vibrate their retinas, recording datapoints from the in-between locations to get in-between pixels.

Re:Still patented too

[ceoyoyo]

Interesting. They're effectively performing super resolution imaging. Unfortunately it has awful noise characteristics, so it's only useful to overcome some real barrier, like your light waves (in a microscope) or your retinal cells (in the spider) being too big.

I believe this is closer to interferometry or synthetic aperture radar.

pixels for mirrors?

[jackelfish]

So essentially, it seems that we go from an array of pixels on the photodetector in the camera, from which the data being collected is filtered and redundant information is deleted during compression into a jpeg. Now a new "less wasteful" method uses an array of tiny mirrors that must turn on and off and then focus the reflected light efficiently onto a single photodetector that then just filters the information using some complex maths? I am not quite sure how this is better, or is it just different?

Re:pixels for mirrors?

[x2A]

"tiny mirrors that must turn on and off and then focus the reflected light"

The lenses are fixed throughout the exposure, so the light's already focuses.

"that then just filters the information using some complex maths?"

Well, the maths actually expands the data to the full size of the image - could be that a million pixels only needs a 3000 readings or somethin.

best of both worlds

[cpearson]

Why could this idea be combined with the current technology. Millions of mirrors AND thousands if not millions of photo detectors would allow faster exposure times without as much waste as current CCD digital cameras.

Windows Vista Help Forum

Re:best of both worlds

[jo7hs2]

Not everyone wants faster exposure times. Default faster exposure times would ruin many shots. Forget motion blur, water blur, etc... Everything would be a perfect freeze-frame. Of course, the consheepers would get razor sharp pictures of their sticky, chocolaty offspring.

Re:best of both worlds

[x2A]

The mirrors can only focus on one place at a time though - to get the light to hit multiple CCDs they'd have to move, or you'd have to move the focus point so the light hits multiple sensors, bluring it, bringing you back to square 1.

Look at the picts

[Var1abl3]

I know it is still in dev but the picture quality they show as examples look like shit... I would never take a pict of my wife and baby with a camera that made them look like a image made out of bathroom tile..... no matter how much battery life it saved.
Also what would this do to a picture of something in motion. I could imagine taking a pict of a car driving by and get the front of the car and by the time the millions of mirrors switch and aim etc the car has passed and I only get a picture of what was behind the car... except for the front bumper or something. Keep working on it... but for now I will stay with my 8MegPix camera and charge the battery everynight.

Re:Look at the picts

[x2A]

"I would never take a pict of my wife and baby with a camera that made them look like a image made out of bathroom tile"

Not all of us have such pretty wives, you insensitive clod!

And the burns my baby suffered from an "accident" we had will actually mean it gets teased LESS in school, and we can actually begin to love it a little.

If it were for these guys....

[d0n quix0te]

We'd still be using a pin hole camera... lenses --wasteful, shutters-just extra parts, zoom -- why would you need it...

-S

Re:If it were for these guys....

[bozendoka]

Focus? Luxury!

Compression

[gilesjuk]

Doh! so what if data is lost in compression? that's why you shoot in RAW format dumbass.

With a 4GB CF card and average RAW image size of about 20MB I don't see any need for JPEG if you have the time to work on the RAW files.

Re:Compression

[Bandman]

I'm torn about this.

There are some shots that I'd like to get that just arn't worth the post processing that raw entails, and they take away from the "good" shots I'm interested in. That being said, I agree with you, RAW is the way to go for any serious photography that doesn't involve film.

Re:Compression

[sahonen]

My camera lets you choose whether to shoot RAW or JPG. If you don't want to post process, shoot JPG, if you do, shoot RAW.

Re:Compression

[cibyr]

Many cameras will let you do both at once :)

Forget about a solid state sensor for each pixel,

[sharkb8]

You can have a million little moving parts in your camera!

The microelectrical mechanical device fabrication techniques used to make the DLP scanning mirrors are taken from tech used to etch transistors. Instead of a circuit bring etched, a movable mirror os etched into slicon or other substrates. And you end up with a bunch of little tiny mirrors moving around on a portable device. Moving parts tend to wear out more rapidly than solid state parts, and are more easily broken. I'd be interested to see how durable this tech is. DLP doesn't have this issue because no one carried a DLP projector or TV around.

Seems like it would have one huge drawback

[Solandri]

Low light sensitivity. Digital cameras gain light sensitivity by acting as light buckets. Moreso for CMOS sensors than CCD, but the important thing is that all those sensor pixels are collecting light for their individual pixel simultaneously - in parallel. With a single pixel sensor, this light collection would have to happen in series to achieve the same light sensitivity. If your shutter speed in low light is 1/25 sec with a 5MP traditional digital camera, in order for a single-pixel camera to take the same picture it would need an exposure time of (1/25 sec/pixel)*(5M pixels)*(10% assumed algorithmic efficiency) = 20,000 sec = 5 hours 33 min 20 sec.

Of course since you're doing all this with mirrors, you could set up a megapixel array and have different mirrors shine at different pixels simultaneously (just like a DLP). But that seems to defeat the purpose of the whole rig.

Re:Seems like it would have one huge drawback

[Eternauta3k]

You could make a single-pixel, b&w camera by putting a very sensitive fluorescent screen between the lens and the mirrors. Expose the screen for your speed, then scan the screen (compensating for fading) with the mirrors.

Re:Seems like it would have one huge drawback

[Antitorgo]

Actually, if you think about it, the minimum exposure time is however long it takes for the mirrors to move and for the CCD to capture the light. With good enough software, motion could potentially be taken into account to remove any blurs and could potentially increase the resolution of the image even.

Since it isn't doing this "one pixel at a time" but instead uses varying patterns to capture the image, the "quality" of the image is dependent on shutter speed vs. exposure of the image. Exposure is just determined in the DSP by the average light levels during and how many samples were taken while the shutter was open. What would be kind of cool is to have a RAW format for this data that could be processed futher "offline" to adjust for things like motion of objects during the shutter-open time and whatnot. The same kind of technology that is used in movie restoration (where they rely on the adjacent images to recreate lost data) could be applied here.

Re:Seems like it would have one huge drawback

[Michael Woodhams]

The one huge drawback is low sensitivity and high noise - two, the two great drawbacks are low sensitivity, high noise and a fanatical devotion to the pope...

Each measurement has noise, and is the sum of a large number of pixels. To determine the value for a given pixel, you have to add and subtract a large number of measurements. Subtracting noisy values increases the relative noise in the result.

Actually, we already have a method whereby a given measurement (pixel) is influenced by a largish portion of the image - it is generally known as an "out of focus picture". Mathematically, it is straight forward to "undo" the blurring, but (as discussed above) inevitably this process magnifies noise.

You might be able to do something interesting with (e.g.) a 10Mpixel mirror array and three 1Mpixel detectors (R, G and B).

Re:Dupe

[orson_of_fort_worth]

Shh. We can score some karma by copying the +5 posts from the original story.

Oh yay!

[malkir]

Sounds like an inexpensive and durable replacement!

Comment removed

[account_deleted]

Comment removed based on user account deletion

RAW mode

[dagamer34]

Or more cameras could shoot pictures in RAW or RAW + JPEG mode. I don't want to spend $300 for a camera that doesn't throw away information when $100 can capture the same info anyway.

Excellent tool for espionage...

[maillemaker]

I have sometimes thought of nano-sized cameras like this that, instead of having a million mirrors to allow a single pixel to take a full picture, instead, only took a pixel's worth of a picture. But each device is like a grain of sand. You could sprinkle the devices where someone is known to be passing through, or sprinkle them on the person, and thousands of these one-pixel devices, working in concert, could generate images.

It would be like "dusting" someone with micro-bugs.

Re:Excellent tool for espionage...

[ShooterNeo]

Right, and in the future, where manufacturing even ordinary objects atom by atom is practical, the future society would have ubiquitous surveillance. EVERY object everywhere would generate ultrahigh res images of everything around it, and nano-scale processors embedded in the objects would analyze the images. Crime would be impossible : even committing a crime would be impossible, as an attempt to commit a major crime would automatically cause an intervention by the AIs watching over everything.

Random?

[Have Blue]

Why don't they just scan over every mirror on the chip in a specific order? Then there would be less complicated mathematics required, right?

Millions of Mirrors * 7 years bad luck = ..uh oh

[JustASlashDotGuy]

'The digital micromirror device, as it is known, consists of a million or more tiny mirrors



Drop this device just one time and you've got bad luck for the rest of your life... or next
million lives if you believe in reincarnation.
I urge all Eisoptrophobia'ist to avoid this at all cost!

I have one... maybe

[dosle]

I'm sure if I looked through my old gadgets box there's something nearing a 1 Pixel digital camera... it also takes 3.5" disks. $10,000 OBO

I have the video camera version

[Anonymous Coward]

This is a short video clip from my wedding:

.................._.....~...._...

Sorry for the shakiness of the video. (And I still can't believe what an ass my brother in law was that day!)

Space/Time tradeoff

[phliar]

If you replace a million sensors with one sensor, for the same sort of exposure you'll need a million times the time. (Or, since the claim for the device is that you don't need to sample everything since you're compressing with JPEG, let's say half a million times.)

But we want the entire frame to be captured in "the same instant" (or you'll see strange artifacts from moving objects).

Let's say we want an exposure of about 1/100s. So, can these micro-mirrors switch at a 5x10^7/s rate (20 nanoseconds)? Since the mirror has to be stable for the interval, the switching time needs to be a fraction of that. So, can these bacterium-sized physical mirrors switch in 10ns?

Re:Space/Time tradeoff

[Negadecimal]

If you replace a million sensors with one sensor, for the same sort of exposure you'll need a million times the time.

Bingo. And to "compensate", you'd have to tweak up the CCD sensitivity to the point of unacceptable noise. I'd also suspect that the power requirements for mechanically moving mirrors would be prohibitive.

Grainy pics, blurry action shots, and a five minute battery life? heh..

Re:Space/Time tradeoff

[GigsVT]

Never heard of MEMS I see.

http://electronics.howstuffworks.com/dlp1.htm

It's fast enough for DLP!

MEMS is one of the major advances in overall technology that the human race has had recently. It's probably as important as the transistor or the laser. Read up.

Your post is a dupe

[Anonymous Coward]

From the more recent mysterious past: http://hardware.slashdot.org/comments.pl?sid=21764 2&cid=17671236

-1 Redundant in 3... 2... 1...

Re:Your post is a dupe

[rumith]

Believe it or not, the post stating that wasn't there when I hit the reply button.

Re:Your post is a dupe

[Anonymous Coward]

Believe it or not

First Post: 01-18-07 04:44 PM
Your Post: 01-18-07 04:54 PM
Difference: 10 minutes

You're right, I don't believe you.

Re:Your post is a dupe

[Aladrin]

"Comment Submitted. There will be a delay before the comment becomes part of the static page."

'nuff said.

Re:Your post is a dupe

[Anonymous Coward]

"Comment Submitted. There will be a delay before the comment becomes part of the static page."

I hope you realize that it doesn't take 10 minutes for Slashdot to update. In fact, that message predates a lot of architectural changes that cause messages to show up almost immediately.

Poor excuse. 'Nuff said.

Scanning back?

[MoxFulder]

I think this design is sort of like an ultra-fast scanning back. A scanning back is a high-end type of digital camera sensor where the sensor has only a very small resolution, but it physically moves and takes a frame at each step. The many resulting frames are then interpolated together appropriately. This can produce EXTREMELY high-resolution images (we're talking 100s of megapixels) but it is sloooow (minutes or hours per exposure). Good for art reproductions and such.

As I understand it, this camera would basically be like a very fast scanning back, because instead of physically moving the sensor for each new frame, the image is changed using extremely high-speed mirrors.

Can anyone who knows more about photographic technology comment on this?

Re:Scanning back?

[modecx]

Actually, this design is more akin to a drum scanner, in my opinion. Drum scanners have a single, very sensitive, very low noise photomultiplier tube, and the tube scans across a rotating drum mechanically. This is like a reverse DLP pointing at a PMT.

Why can't we have an array of these, kind of like a Beowulf cluster of 'em? Seems to me that if one is good, a thousand must be better.

Demo here

[Anonymous Coward]

Here's the image:

.

If you can't see anything more than a single gray dot, it's because your monitor doesn't support moving the image randomly and rapidly enough.

HDR!

[ACMENEWSLLC]

What is cool about this is that it could allow HDR(http://en.wikipedia.org/wiki/High_dynamic_rang e_imaging) in the camera itself.

While you eye can see many different luminosities of light, a camera has limited contrast. Since it is taking not a single picture, but millions of them in an instant - it could also adjust contrast dynamically.

That would be cool.

Re:HDR!

[ceoyoyo]

You can perform HDR in a conventional camera much more easily by having it take two or more exposures with different aperture or shutter speeds. Mine does that just fine, actually. It just lacks the software in-camera to put them together.

Re:HDR!

[Firehed]

Yeah, I can do AEB too. Up to three different exposures easily, though I can of course re-adjust things and do it again. The difference here is that it would properly expose each of the six million pixels-to-be and automatically stitch them together, thusly avoiding the need for Photoshop.

I'd be inclined to think that such a thing being performed automatically within the camera for every shot taken (at least potentially) would be much easier than fiddling with features to take some AEB shots and then using a Merge to HDR feature in Photoshop (or whatever software you prefer). But if six steps is indeed easier than one, do let me know.

Of course, I don't see how this avoids "wasting" data any more than a traditional camera would. You either compress the image and "waste" that data again, as any point-and-shoot does, or shoot in a micro-mirror equivalent of RAW, and keep all of the original data. It would seem to me that rotating around millions of bacteria-sized micromirrors to take a wonky pseudo-RAW shot would be a lot more complicated than just flipping up a single larger mirror as any dSLR does today.

Re:HDR!

[ceoyoyo]

You could do HDR all in the camera with nothing to fiddle with just as easily (probably more so) with a regular camera though. It's not implemented in any camera that I know of, but it certainly could be.

To do it with this camera you'd have to... well... I'm not quite sure. Fiddle all the mirrors once and then do it all again faster and slower I guess. And if your random fiddling pattern from the first time doesn't match up well with the second time you have to do some interpolation, which will add noise to your image.

Re:HDR!

[ACMENEWSLLC]

If you take 3 shots and do HDR, you have the contrast of your camera (500? 1000?) times the number of pictures (3.)

Doing this on each pixel lets you have unlimited contrast. The human eye can see something like 50,000 levels of contrast. I can't find the document stating the specific number. But it's a lot more than what 3 HDR images could do.

Thing of this. You could take a scenery picture in the middle of the night, capture the stars, the moon in detail, the moonlight reflecting off the lake, the street light in perfect clarity, and the entire forest in full detail because you are adjusting for each pixel. You could change the exposure time for each pixel.

I think if someone were to actually do this and show an HDR picture side by side (printed) you would be amazed. Most monitors are not capable of displaying this amount of contrast.

Re:HDR!

[ceoyoyo]

People (including me) do HDR all the time. It's useful in certain circumstances where you have a lot of contrast, but often produces a somewhat unreal feeling.

I think you misread my post. The point was that a one pixel camera doesn't let you do HDR any more easily than a standard camera.

Here's the procedure. One pixel camera: make all the measurements you need for a regular exposure, vary exposure settings, make the same measurements again, vary exposure settings, make the same measurements again.

Regular camera: make all the measurements you need for a regular exposure, vary exposure settings, make the same measurements again, vary exposure settings, make the same measurements again.

Same thing in both cases. HDR is not an advantage of a one pixel camera.

dumb dumb dumb

[cwm9]

I can't see any application for which this is an advantage, not even astronomy. In astronomy I guess you could argue that you want a highly efficient detector because you need to harvest every photon you can... but a larger detector also means a larger dark current, so that seems to be going the opposite direction you need to be going.

For any other normal application this has got to be the stupidest thing I've read in a while.

This device will throw away >99% of the available light. So instead of shooting f8/125 portraits you'll be shooting f1.4/4 blur-blobs. Yeah, that's what every photographer wants.

The basis

[Jerf]

Here's some of the basic math behind the idea:

When you lossfully compress an image, you are literally throwing away data. If you compress a 1MB image down to 100 KB, which with JPG is still very good quality, you are mapping many, many, many slightly different but ultimately very similar source images all onto the same compressed image.

Consumer cameras "waste" time starting from a full lossless image, and compressing it with JPG; the waste comes from collecting all of this data that has no bearing on the final result. (Anything that stores the .RAW of the image isn't doing this compression, it's storing the entire data set.)

The idea of this system is that by mixing the pixels together in a certain way, we can collect less information in the first place. For what would be a 1MB picture in a standard camera, you'd start off by only collecting 100KB of information, and then computing the image from your sequential numbers.

Two problems leap to mind:

• I find it very, very hard to believe that "random" is the optimal approach. I would have thought there would be something much better than that for the bases, but I could be wrong. (There almost certainly is something better than "random" but it may not be better enough to justify the computational expense.)
• JPG bases were carefully designed to match the human visual perception system and make it difficult for us to perceive the compression artifacts. The compression bases in this situation will have to be optimized for information gathering, which won't be the same as the human eye, which will result in somewhat inferior pictures, bit for bit. If you know what you're looking for, you can see it in their sample pictures; it's going to take a lot more bits to make that mosaic effect "go away" that it will to make JPG artifacts "go away".

A clever PhD may be able to solve both problems in one swipe, by using a clever mirror progression that happens to map better to the JPG standard. (You can't get it perfect though because you can't predict in advance how many bits go to one JPG block, that's computed dynamically.)

It works, and it's a clever algorithm, but I would definitely still question its practical usefulness over a conventional imaging system. I think the current trend of compression is temporary; the megapixel race should start to slow down (who needs 100megapixel pictures of their baby?) and then as cameras and storage continue to advance, we'll start getting uncompressed or losslessly compressed images instead. I could see this technology winning the race to be the first to produce a single camera that matches the image capturing power of the human eye, though; by manipulating the incoming light you may better be able to manage widely varying light levels.

(Finally, bear in mind before posting criticisms of how impossible this all is that they appear to have actually built a device that does this, which trumps skepticism.)

Re:The basis

[ceoyoyo]

They can do it, but under extremely controlled situations. Lots of image processing techniques work great until you throw a bit of noise or movement into the mix and then things break down.

Re:The basis

[evangellydonut]

JPEG2000 has a lossless algorithm definition...

the other thing that bugged me about the article is about the "waste of power to toss away information" is pretty much awash when you factor in the math used to piece together the mirror reflections vs raw-to-jpg compression, which ultimately are all done on some sub-90nm CMOS ASIC that consumes most of the power through idle current leak rather than peak utilization...

from an amature photographer pov, it's totally useless as I'd rather keep the 12MB raw file and perform long and arduous DSP on it via software later, since DSP algorithm will improve but my camera's hardware is fixed... but that's a different debate.

Re:The basis

[tfinniga]

IIRC from the original story, or one of them that came out, this isn't really much of a win for the average digital photographer. Where it really shines is when your sensors are extremely expensive - say, actually recording individual photon wavelengths instead of just filtered intensities, or something esoteric like that.

So, if an array of sensors is economically unfeasible, you can instead have an array of mirrors and a single sensor.

As for your question about sampling methods, there are several feature-sensitive sampling methods in use in high-end rendering (Monte Carlo ray tracing, Metropolis light transport, etc). These are pretty computationally expensive and/or complicated for a consumer device. And if you're using an esoteric sensor, you might be doing scientific work, and want an unbiased sample anyway.

Re:The basis

[Anonymous Coward]

this system seems ideally designed for satellite imagry....
-the patent seems to point out how it removes some noise from turbulent atmosphere
-they squish down the power on the front end transmit the raw sensor info, and process it in the back end.. that gives you more power to move the sattelite around , or take out weight from power generation to make it lighter and easier to launch...
-if you maintian the pixel chipping pattern as a secret key it comes encrypted for free! a huge 2-D Linear feedback shift register!

although providing a coherent source would be pretty tough....

Re:1MP?

[Speed Pour]

More like a 3000 core processor clocked at 1 mhz

Damnit, I can SMP like nobody's business!!!! just really slowly :(

Re:1MP?

[speculatrix]

no, a beowulf cluster of 1 bit processor core CPUs.

mod parent down!

[Achoi77]

freakin goatse trolls!

Re:mod parent down!

[Khyber]

With a goatse troll posting . as their wife's bum, I'd have to feel really sorry for the guy that could actually penetrate that without killing her!

cheaper?

[burnin1965]

Are micro-mirror arrays truely cheaper to produce than a CCD or CMOS photo sensitive array?

fft

[Anonymous Coward]

theoretically it should work with one pixel, the light in the pinhole camera also passes in one point (theoreticaly). In this point all the information is stored in the frequency domain. But you don't need any mirrors just an really tiny sensor.

Pixel

[postmortem]

Sigle pixel on my laptop screen looks pretty small.. I'm not sure that I'll be able to enjoy my pr0n sessions as much with this camera.

Re:1MP?

[Ambush Commander]

No, no, no, not one megapixel, 0.000001 MP! (If your magic is that low, you probably need to restock on pots.)

Or how Wired Magazine will report...

[Anonymous Coward]

On newstands soon:

* Gigapixel Cameras On the Horizon

[* = "Nano-" in rather small letters.]

When Nanotechnology Meets Digital Photography.

[picture: closeup of a laboratory microscope, with a professional camera -- ostensibly of microscopic size -- visible through its eye-piece]

Do you think your ten megapixel camera is teh awsom? Well with nanotechnology, 1,000 megapixel (no, that's not a typo, that's one thousand megapixels, that is, a gigapixel) cameras might be just around the corner -- 1,000 nano-megapixel cameras that is. For more about the coming invasion of the "nanogigapixel cameras", and the bottom line for your wallet, turn to Page 27 for our cover story.

uh-oh

[Protein Geek]

so if i understand this right, each mirror has to keep working to get a good photo. so every time a single mirror in the array screws up, a pixel that gets compiled to a photo gets screwed up? yippee! 35mm SLR 3

Single pixel?

[dal20402]

The QuickTake is back.

Re:Forget about a solid state sensor for each pixe

[Anonymous Coward]

You've almost certainly been carrying a similar moving part around with you for some time with no problems. The quartz crystal oscillators used to generate the clock frequencies for portable electronics (including some cell phones) are electro-mechanical parts, and it takes quite a bit to get them to fail. In fact, I don't think I've EVER had a crystal fail, and I'm pretty rough on gear.

Better idea?

[Anonymous Coward]

How about a fiber optic array positioned as a lense - one fiber per pixel - where every fiber has a different transmission speed (could be done just with length?)

So then, each fiber transmits the pixel to the CCD at a slightly offset time.

Sounds cheaper, more durable and easier than timing and focusing a bunch of mirrors?

Exposure?

[dangitman]

Modern cameras have sophisticated algorithms to determine automatic exposure based on the composition of light and dark within the frame. To get the proper exposure with this system, it seems to mean you'd have to pre-scan the scene to get the exposure value. Or go back to old-school external light sensors rather than TTL (Through The Lens, not Transistor-Transistor Logic) metering.

Re:Exposure?

[GigsVT]

On the contrary, one of the drawbacks of normal cameras is that you can only have one exposure level for the entire image.

I saw a lecture on a technology that takes 3 different exposure pictures, then combines them to put a lot more depth into a picture with some bright areas and some dark areas.

I don't know if this has made it into hardware yet, but with this serial method, each pixel could have a different exposure.

Think outside the box man! The idea that you only get one shot at an exposure level is a throwback to analog film.

Re:Exposure?

[dangitman]

On the contrary, one of the drawbacks of normal cameras is that you can only have one exposure level for the entire image.

Why does that have to be so? Couldn't you have different exposure levels for different pixels coming from an array if it were designed that way?

I saw a lecture on a technology that takes 3 different exposure pictures, then combines them to put a lot more depth into a picture with some bright areas and some dark areas.

A technology? You can do this with normal cameras. I've been doing it for years. You can also use a similar technique to reduce the noise from digital cameras and scanned images.

It's hard to see it as a "technology" as such, when it's a simple photographic technique that has been used since the very early days of chemical photography. It's not some "new thing" that is unique to digital photography.

I don't know if this has made it into hardware yet, but with this serial method, each pixel could have a different exposure.

Again, how is this technique specific to single-pixel cameras, when it could just as easily be applied to pixel arrays? And what happens if the camera moves between the exposure reading and the actual exposure?

Think outside the box man! The idea that you only get one shot at an exposure level is a throwback to analog film.

When did I say you only have one shot at an exposure level? Jeez, analog photographers have been using multiple exposure readings since the days of yore.

Re:Exposure?

[GigsVT]

It's hard to see it as a "technology" as such, when it's a simple photographic technique that has been used since the very early days of chemical photography

How? Masking off parts of the image? This algorithm combined each pixel in the image to get greater depth. It is not some simple cut and paste photoshop trick.

when it could just as easily be applied to pixel arrays?

I don't think easily would be the right word. You would have to turn of some pixels before others, to get varying exposure on the same image. The only other way to do it would be the way it was demonstrated, by taking 3 distinct images, which doesn't work so well on a moving target.

old news

[Anonymous Coward]

Come on, I read this months ago http://www.usatoday.com/tech/science/columnist/200 6-09-30-single-pixel_x.htm

A tremendous breakthrough!

[meanfriend]

Single pixel images will revolutionize the efficiency of porn sharing.

Are you into hentai? Here you go! .....................

Barely legal teens? Coming right up .......................

Even goatse freaks dont need to be left out:
.

Though I'll probably get modded down for that last one :(

But there are lots of mirrors pointing

[Anonymous Coward]

Lots of mirrors at one time means plenty of light.

Not quite...

[Anonymous Coward]

Nah, the dupe ??? comes before the dupe profit.

Sorry :-)

The Obligatory...

[Jon Eiche]

Puts your two-bit camera to shame!

So...

[Anonymous Coward]

Instead of spending dollars and seconds on a bigger CCD, I'm spending dozens or hundreds of dollars and minutes or hours compiling thousands of pixels together into a single picture?

Yeah, REALLY stretch to save me some effort there researchers, thanks, assclowns.

The new Flintstone Digital Camera

[iminplaya]

Instead of just one bird making the whole picture, we now have a million all waiting in turn the mark the part they saw. Millions of rapidly moving parts is not exactly my idea of reliable. I thought the idea of solid state was to have no moving parts to wear out.

good for video

[r00t]

If you record the pixels in random order, you mostly free yourself from having a specific interleave and framerate. This is especially true if the "pixel" is single-color.

Having freed yourself, you can easily produce output in different framerates with/without interleave. Normally, format conversion causes various bad artifacts. With this, you could easily output any framerate (24, 25, 29.97, 30, 48, 50, 59.94, 60, 72) you desire.

great profit engine

[mgabrys_sf]

To capture a 4 megapixel image with these cameras, I'd only need to buy 4 million of them. BRILLIANT! Talk about a sales coup!

This is related to Dual Photography

[Snorlax]

There is an interesting related subject called Dual Photography, where the "light source" is one (or more) simple photocells, and the "camera" is a video projector or steerable point light source such as a laser pointer.
The computer can render what the image would look like under any combination of lighting from the original photocell position(s). This technique can create images with just a point light source and a photocell, and no camera or lens at all.
This is only applicable to "still life" photos, where the subject sits still in a darkened room for several minutes while the light source scans the scene. The linked page summarizes the technique, and links to a 19MB PDF paper that contains more info and several example images.

Single-pixel cameras been around for decades...

[Anonymous Coward]

Single-pixel cameras have been around for decades... they're called "photocells" .

Re:Single-pixel cameras been around for decades...

[Anonymous Coward]

Whoosh! went the joke.

Technology is moving so fast...

[Curate]

I've been following the digital camera industry closely for the past few years. We had 3 megapixels, then 4, 5, etc. Now they're blowing that out of the water with a 0.000001 megapixel camera!!! Amazing! There's no frinkin way they're topping this, baby! I guess with ongoing advances in miniaturization, maybe someday they'll find a way to cram 0.000000001 gigapixels into a camera. Today, such a camera would be the size of house.

Re:Technology is moving so fast...

[DemonThing]

I suppose after that some clever guy would come up with a 0.0000001-megapixel camera.

great

[Kanasta]

instead of having a million perfect pixels
now we need a million perfect tiny mirrors

One Dead Pixel!

[BenJeremy]

Just my luck, and the warranty says I can't return it unless I find at least 4 dead pixels!!!

Old idea on new equipment.

[Anonymous Coward]

The JPL/NASA Pioneer 10 & 11 used this method to get their images from Jupiter and Saturn. Also the Viking 1 & 2 used this method to get their image of Mars. They each had "high quality" sensor (at the time of manufacture) so they used a mirror to scan the image to the single "high quality" sensor.

Thousands of individual snapshots huh

[HaMMeReD3]

If there is less then 1 million single pixels captured, I don't care about noise ratio or anything like that. We are talking about low-resolution photos here. Although I see the potential here for recreating potentially infinite resolution, time is a factor. This almost reminds me of people who convert scanners into digicams. The pictures clearly show that there was a large delta in the time it took to take the image, not really in exposure time, but in each scanline, so you get really wierd warping effects.

What I'd like to see more then this is a 6+ megapixel camera that's smart enough to turn one long exposure into a hdr image. It could do so by taking "snapshots" of the photo while it is still processing, so a 2 second exposure would have snapshots taken at 0.5, 1, 1.5 seconds, then it could compare the photos to create response curves and produce a HDR output. Anyone who has played with photoshops merge to hdr functionality would know how well it works.

Speed of all this?

[MattS423]

Shutter speed? My camera takes some time to take a picture...particularly in dark environments where I can't use the flash (Fireworks for instance). These pictures usually come out blurry because in the .5 second it took to take the picture, my hand moved. If this thing has to record say, 5 million pixels (a 5MP camera), surely that takes some time. Is it long enough for me to move my hand and make the picture blurry? (or, otherwise askew?)

Economist article

[Anonymous Coward]

I just read about this in the Economist.

Sounds great and all, but there are a few minor drawbacks: it's the size of a table and takes ~5 minutes to take a picture.

what if...

[plbg32]

what if we used this tech with say a 10 meg camera, would that say give me something like a 1000 meg pic?

NASA used one on Mars nearly 30 years ago

[Anonymous Coward]

This is not new.
Remember those first stunning panoramic pictures from Mars decades ago?
They were taken with a single-pixel camera.
NASA used a single-pixel camera on its Viking Mars lander, which launched in 1975!
http://history.nasa.gov/SP-425/ch3.htm

Last time I checked,

[NalosLayor]

CMOS and CCD sensors were already pretty damn cheap, and even low end consumer grade cameras had as much as 7 Megapixels. The expense in high end stuff, I believe, is in the optics, storage systems proprietary logic and the like. Not to say, of course, that all sensors are the same, but still...

Go to the Mirror, Boy

[Doc Ruby]

If they replaced millions of electronic photodetectors (photodiodes, CCDs, whatever) with a single cheaper, more consistently colored mirror, that they micropositioned millions of times to scan an image extremely quickly, without suffering the skew distortion from vibrations while scanning, then this could be a much cheaper, more consistent imaging device, Maybe much much smaller, and maybe getting much higher yields than large photodetector arrays which lose parts to defects per mm^2. But is an array of MEMs mirrors really cheaper than electronics?

Seems like it couldn't work for a consumer product

[Anonymous Coward]

Why wouldn't it work?

Because I think it'd be a VERY long time before they could get that thing to sync up to a flash.

The strobe on your average disposable camera is somewhere between 1/1000th of a second and 1/10,000 of a second.

I don't think the article mentions how long it takes for these micromirrors to collect 1MP worth of data, but if they had to do it in 1/1000 of a second, we're talking about what, 1 nanosecond per pixel?

Seems like it might be a while before that's possible... Still a cool idea, though.

The Bill Gates-Land Camera...

[Teddy Beartuzzi]

1 pixel ought to be enough for anybody!

/. at work

[Kuvter]

I read Slashdot at work, come on have some decency or at least forewarn us of stuff like that.

There's no such thing

[ear1grey]

"Complex mathematics then interprets the signals..." Swoon! Thats complex mathematics, the kind that we plebs cannot hope to grasp in our tiny little minds. I note also that this is mathematics so complex that it has achieved sentience and now interprets signals itself, no longer needing a computing machine to help. Time to sell INTC and AMD?

No shit

[Khyber]

1. Get original entry to pique interest.
2. Post duplicate a couple years later.
3. Gain revenue from dupe story.
4. ???
5. Profit EVEN MORE!!!!

Hasn't this been proven, already? See, the reason TAGS are here is so we can show this has been posted before. Such and such story comes up. Well, it can only have so many tags. Check the tag matches, and whammo, you've got a narrowly-defined list of good matches that could indicate a dupe. Add in a quick bit of code to check links, and maybe search by itself thru those links for anything that shows the submission as a duplicate, such as other links posted before.

And to the person who mentioned submitting slashcode and then the other person talking about thinkers not having experience - that's what thinkers are for. We come up with ideas. You engineers that do it for reality try to figure out a way to make said idea work. The problem ends up being you're to hesitant to tell us we're out of our minds, so stupid things get thru, and money gets lost, and people lose their jobs.

ENGINEERS NEED TO BE MORE VOCAL DAMNIT! Band together and go against your supervisors. If they fire you all, they'll LOSE. Stand up and tell them what the fuck they need to know, and why it must be done this way! SCREW THEIR DEADLINES!

the only person that knows enough about the business to say a goddamned thing is the person working their ass off at the base level, who has to understand EVERYTHING for the most part in order to be able to meet the quality expectations of the company. They know the flaws and weaknesses better than ANYONE ELSE. They hold the answer and key. Ignore them, and risk getting a shit product with poor quality out of the door to consumers. Even the repair technicians can tell you what the fuck's wrong with your product (this assumes electronics manufacturers, BTW.)

early video cameras were 'single pixel'

[gp310ad]

the only differece here is the scanning mechanism.

Tha flying spot scanner and Farnsworth's image dissector are both 'single pixel' devices.

Isn't it compression

[Joebert]

What's the difference between a million pixels, & a million mirrors ?
How will this deal with highspeed photography ?
Will the image become blurred if for some reason the rate at which the pixels change is altered by a fraction of a second before the entire array of mirrors is read ?

Re:Stupid is as Stupid Does

[cbacba]

This 1 pixel camera is one of the most stupid ideas I've ever encountered and shows a total lack of comprehension on the topic.

First off, all those mirrors are mechanical things which require control circuitry and mechanisms and complex mechanical design - at least compared to the a piece of flat silicon with something etched in it. And, to some extent it seems like someone looked at a complex cmos area sensor and shouted "eureka I done found it - we could invent a simpler gizmo - like maybe take a sheet of paper and curl it up on a cylinder, spin it and run a light and photocell along the top and scan that sheet of paper and send the scan over the telephone line".

What's worse is that light consists of photons, each with around 1-2 eV of energy so they are quite discrete at visible wavelengths. That means at a given light intensity level that to actually get a photo image, it takes a sensor of a certain size so much time to aqcuire enough photons to register its relative intensity. That must be done for each pixel of the final image. If the time required is 1 millisecond per pixel - then for an 8 megapixel image, it's only gonna take 2 hours to capture the data - in which nothing better change. I'm sure the kids will be happy to sit still and hold their breaths that long.

In computer terms, this is the equivalent of looking at a future Cray parallel processing supercomputer, shouting eureka i found it, and then proceeding to (re)invent the turing machine.

The use of mechanical systems to achieve image capture (or use to achieve higher resolution images) has been around for decades. The original tv camera used a series of holes in a rotating disk to achieve a serial stream of video data. When ccd arrays were small, expensive and with lots of defects, linear ccds were used with rotating mirrors to achieve higher resolution.

In fact, with modern processing, adding in random vibrations (directional variations) into a camera mount can be used to achieve higher resolution images from a video stream than the pixel count of the sensor.

The notion of digressing to what looks like a much more expensive technology to achieve vastly limited and inferior results to existing technology makes me wonder why these guys have yet to be awarded the darwin award of the century. Maybe that's because they didn't know to plug in the lamp when they were sticking their tongues in the empty lamp socket.

Japanese do it first

[jespejo]

...just a little simple implementation http://rgby.jpn.org/product/index.html

For those blank wall pictures

[Chris Shannon]

For example a picture of a blank wall will have many pixels with the same colour and texture information.
Dr Baraniuk said that this is where the single-pixel camera really has an advantage.
The perfect camera for those ultra mundane pictures!

Re:1MP?

[gstoddart]

no, a beowulf cluster of 1 bit processor core CPUs.

As opposed to the 2-bit CPUs which make up the average Beowulf cluster? ;-)

Less random than random

[GlobalEcho]

Perhaps the sampling would go better with a quasirandom sequence.

Tags

[bcmm]

Is "." a valid tag?