What is Digg?21 Comments
- trogdor282, on 10/12/2007, -0/+10I would say it's much higher res than any other single-pixel camera ;)
This is just a proof of concept. It sounds like the crazy math they discovered is more significant than the device itself. Eventually it will lead to new things that are actually useful. - ABadInAlbany, on 10/12/2007, -1/+8the technological breakthrough the porn industry has been dying for!
- somespecialist, on 10/12/2007, -0/+6I'm in Kelly's Electronic Materials class at Rice. The guy knows more about nanoscale physics than anyone at the university. If you're interested in his work on this camera, you'd probably be interested in his work on "nanocars." Cool stuff.
- wesball, on 10/12/2007, -0/+4You know computers were once these incredibly slow, archaic machines that took up entire rooms. At the time, no one had any idea what impact they would have on the world.
I wouldn't be so quick to scoff this.
Imagine a world covered in millions of these microscopic cameras, all networked together, creating a virtual environment of almost limitless detail... - masamunecyrus, on 10/12/2007, -0/+3Pictures of it in action:
http://www.dsp.ece.rice.edu/cs/cscamera/
Basically, instead of capturing a 1-megapixel image tons of times, and then throwing out most of the data, this thing captures the image pixel-by-pixel, only once each, thus reducing power consumption and reducing waste.
At least, that's what I got from its explanation. - mhockey14221, on 10/12/2007, -0/+3If this idea is possible, it would take DECADES for it to advance to that level. Did you see those 128x128s?
- geminitojanus, on 10/12/2007, -0/+3"I'm kinda thinking outside the box here, but this technology could mean that we don't have to cram so many pixels onto a tiny sensor, giving it much higher dynamic range since it won't be as sensitive. This could also mean that for the same size sensor, we could put less pixels but end up with the same resolution, giving us greater depth of field, eventually leading to compact point-shoot cameras that rival DSLRs."
The problem is the Digital Mirror Device (DMD). To increase the resolution, you need to increase the size of the constructed DMD. The DMD is a nano-device that has a bunch of tiny mirrors that are turned on and off by a signal (in this case from a random noise generator). Making these devices is expensive. It's controlled a lot like an LCD, only on a silicon chip (but it ISN'T LCoS, which actually IS a liquid crystal on silicon), if you can imagine that. That cost is reflected (PUN) in the cost of DLP tvs. And essentially, that's all this system really is; a DLP TV in reverse (using mirrors to reflect the image into a photodiode instead of using mirrors to reflect light from a lamp through a color wheel). The photodiode is sampled at various states of mirror orientation and the output is amplified and sampled by an A/D receiver.
This is where the math kicks in, using a lot of math I'm not familiar with at all (which is sad, because all of this wavelet stuff is getting really popular), it reconstructs the image using a number of recorded voltages in correspondence to the mirror positions (one of which is the base position of "off", where all of the mirrors are tilted away from the light source and the DC offset can be recorded from the photodiode to compare against the other sampled voltages). [ Note: If you can, get into this class: http://www-dsp.rice.edu/courses/elec631/ , as this is where they're teaching "Compressive Sensing", the foundation theories behind this new format].
The interesting part is that the image from this device would correspond both to the positions of the mirrors, and simply be a measure of many voltages with high precision. That means the image is fairly compresses as it comes in (instead of being the measure of literally millions of voltages in the case of a one megapixel camera, this is the measure of 5,000 or less voltage values). It also means that the post processing becomes a huge part of how good the picture looks, and that the best way to make pictures look better from this camera are to suppress noise in the circuit, and use a high quality, high bitrate ADC (the faster it can sample, the faster you can take pictures with this camera; it wouldn't be unimaginable to be able to build a digital camera to take hundreds of frames a second, though the hard part will be knowing where to cut a frame, as the image is likely to change very slowly within hundreds of ADC samples).
What I'm getting at is that this camera won't take as good of pictures as your current digital camera (not for a long time anyways), but it's likely to take a lot more of them, and be able to store more of them on the memory card, and it should use less power. But, because of the optics involved, because of the DLP chip, and because of the ADC needs, this camera will be expensive, and will likely be best served on a space probe (at least, for the next 20 years, until it breaks the price floor for consumers like our current digital cameras did). It'll also be very impressive to see what this kind of sensing will be able to do for Scanning Electron Microscopy(CS algorithm)/Scanning Photo Microscopy (camera implementation). - Livewire, on 10/12/2007, -0/+3I'm kinda thinking outside the box here, but this technology could mean that we don't have to cram so many pixels onto a tiny sensor, giving it much higher dynamic range since it won't be as sensitive. This could also mean that for the same size sensor, we could put less pixels but end up with the same resolution, giving us greater depth of field, eventually leading to compact point-shoot cameras that rival DSLRs.
Basicall, we could one day have the quality, dynamic range, and depth of field of a camera this size:
http://images.digitalkamera.de/Test/MN-NikonD70-Front-rechts-Blitz-ML.jpg
in the body of a camera this size:
http://www.nestor.minsk.by/kg/news/2006/07/2606.jpg
DUGG - donjaime, on 10/12/2007, -0/+2Its still in development and its an interesting idea. But I can see some drawbacks already. The optics will be decidedly more complex. And more importantly, the exposure times would go way up when compared to a standard CCD camera. Forget trying to take a picture of a moving object with that. Recording video would be even harder to do.
- geminitojanus, on 10/12/2007, -0/+1The DMD is there to act as a contrast material for the reflected image; by taking the difference between the DMD and the field, you get an accurate read on the image. By increasing the complexity of the noise (but keeping the amplitude the same, not something you have to worry much about with light), you can increase the quality of the image you reassemble. Rotating or shaking the device would only make the picture blurry, as the contrast would be incomplete within the "pixels".
- ardenr, on 10/12/2007, -0/+1I want to live long enough to see this made into a hi-def video camera.. not my reason for living though. out of curiousity, what could you achieve with a rotating DMD? or gyrating?
- drenader, on 10/12/2007, -1/+2A mega pixel is 1 million pixels, so this camera using only 1 pixel to take an image that is even distinguishable is monumental.
- donjaime, on 10/12/2007, -1/+2What would be cool is if they created an ultra high resolution camera by combining the stochastic sampling techniques they are using with a standard high resolution ccd.
Damn... that would be awesome. I wonder if some camera manufacturers are already doing something like that... - gaijin, on 10/12/2007, -1/+1@somespecialist
"The guy knows more about nanoscale physics than anyone at the university..." That would put him somewhere in the top 2 percent in the world. Go you Fightin' Rice Owls! - nreynolds, on 10/12/2007, -2/+2it's equal to about a 1 megapixel shot. not high-res in my book.
- snorb, on 10/12/2007, -0/+0I'm familiar with the math referenced in the article, and the terminology used is a little misleading. If by pixel, you mean a small square area of the image, then camera is not measuring a single pixel at a time, which would just be stupid. What it is doing, is measuring the sum of the values of a random subset of pixels. Another way to think about it is you're taking the inner product of the image with a random vector of 0's and 1's. I guess you can call this measurement a pixel, but I think it's kind of unintuitive. But I guess the phrase "single-pixel camera" gets the message across of how revolutionary this is, more than "random-subset-measurement camera".
- Ouroboros, on 10/12/2007, -0/+0I wonder if there is any relationship between this and the Red Mysterium sensor.
- NotSoBright, on 10/12/2007, -0/+0This is where one decides which will be easier to produce, be less prone to noise, and will scale better. A physical nano-mirror, or the size of say transistors?
- Wanon, on 10/12/2007, -10/+0In fact with only 1 pixel, I would say it's quite low resolution...
/has not read tfa - shortarabguy, on 10/12/2007, -12/+0Uh... So it's a 1 megapixel camera that takes about 5 minutes to capture a single image?
I'm sorry, but I'm perfectly content putting down $200 for my "inferior" conventional, 6 megapixel camera...
And yes, I'm fully aware that it's made for scientific purposes, but they did mention adapting it for "regular," consumer use, which seems to be a bit unrealistic considering that the benefit will be minimal, as far as I can tell.
© Digg Inc. 2009
— Content created and posted by Digg users is 
