from the if-its-not-daguerreotype-it-might-as-well-be-photoshop dept.
dotdotdot writes:
An extremely tiny lensless camera (PDF), developed by Rambus, has been slowly making waves over the past year. Researchers for the company, David Stork and Patrick Gill won a Best Paper award at last year's Sencomm 2013 for describing what the company has created.
We describe a new class of lensless, ultra-miniature computational imagers and image sensors employing special optical phase gratings integrated with CMOS photodetector matrices. Because such imagers have no lens, they are ultraminiature (~100 µm), have large effective depth of field (1 mm to infinity), and are very inexpensive (a few Euro cents). The grating acts as a two-dimensional visual 'chirp' and preserves image power throughout the Fourier plane (and hence preserves image information); the final digital image is not captured as in a traditional camera but instead computed from raw photodetector signals. The novel representation at the photodetectors demands that algorithms such as deconvolution, Bayesian estimation, or matrix inversion with Tikhonov regularization be used to compute the image, each having different bandwidth, space and computational complexities for a given image fidelity.
Such imaging architectures can also be tailored to extract application-specific information or compute decisions (rather than compute an image) based on the optical signal. In most cases, both the phase grating and the signal processing can be optimized for the information in the visual field and the task at hand. Our sensor design methodology relies on modular parallel and computationally efficient software tools for simulating optical diffraction, for CAD design and layout of gratings themselves, and for sensor signal processing. These sensors are so small they should find use in endoscopy, medical sensing, machine inspection, surveillance and the Internet of Things, and are so inexpensive that they should find use in distributed network applications and in a number of single-use scenarios, for instance in military theaters and hazardous natural and industrial conditions.
(Score: 3, Funny) by e_armadillo on Friday March 28 2014, @09:48PM
Not just existing as a patent troll? Now that is news!
"How are we gonna get out of here?" ... "We'll dig our way out!" ... "No, no, dig UP stupid!"
(Score: 2) by edIII on Saturday March 29 2014, @12:05AM
HA HA Ha
That's what worries me here. A company like Rambus with its litigation history, delusions, etc. gaining power.
They are not a nice company. They are all some lying bastards...
Technically, lunchtime is at any moment. It's just a wave function.
(Score: 2) by davester666 on Saturday March 29 2014, @03:31AM
It will be licensed on a pay-per-image basis.
(Score: 3, Interesting) by bob_super on Friday March 28 2014, @10:08PM
Star Trek had it wrong.
The far future may be about people stepping in a space and getting a fake world projected around them from computers.
The near future is people walking around in a world in which computers record absolutely everything around them.
I'm sure people with Old Timer's memory issues, as well as Indian women, may appreciate the progress.
The rest of us may want to work on wearable mini-EMPs (camera-be-gone)
(Score: 5, Informative) by dx3bydt3 on Friday March 28 2014, @10:32PM
The replacement for the lens in this case is the pattern etched glass whence the sensor records transmitted light. If this optical system behaves like a lens does, then the diameter of the pattern etched glass would be analogous to the aperture of an objective lens. The article states that the etched glass and chip are the size of a period. So, if the internet is to be believed a period is about 0.3mm diameter. Based on Dawe's limit [wikipedia.org], this would translate to an angular resolving power of 387 arcseconds or about a 10th of a degree. To put that in perspective, if this camera system works the way typical cameras do, it would only be able to resolve a 6mm object from a ~3m distance. I wonder if this approach yields better results for its size?
(Score: 4, Interesting) by cosurgi on Friday March 28 2014, @11:12PM
That .pdf presentation is extremely interesting.
The primary approach is using Fourier transform to reconstruct the image which passed though a diffraction grating specifically designed so that at all (visible) wavelengths and at all incident angles the Fourier transform does not yield zero, which makes anti-convolution possible. Very important point here is using destructive interference (opposite phases cancel each other) at different places on CMOS matrix, for light coming from the same incident angle. This allows using Fourier transform, and wouldn't be possible without grating. Very interesting method. It is the "next step" which you make after you realize that the ordinary lens is actually doing a Fourier transform on the focal plane.
The CMOS sensor is roughly 0.8x0.8mm.
The depth of field is from 1mm to infinity.
The method used can also computationally correct any lens aberrations (if you use lens in front of diffraction grating). You only need to calibrate once, so that the Fourier transform of aberration of these lens is recorded.
I see that this method is a wave of the future. Mostly because it is much cheaper than traditional optics, and also is smaller. In few years this will be on every phone.
I don't see any patents. Anyone cares to check if this is free research, or did they patent that?
PS: what is Dawe's limit that you mentioned?
#
#\ @ ? [adom.de] Colonize Mars [kozicki.pl]
#
(Score: 5, Informative) by dx3bydt3 on Friday March 28 2014, @11:45PM
Dawe's limit is the limit of resolving power of an optical system, which is limited due to diffraction. I perhaps could have linked angular resolution [wikipedia.org] instead, or diffraction limited system [wikipedia.org].
(Score: 3, Interesting) by anubi on Saturday March 29 2014, @12:52AM
I am particularly intrigued by the depth of field as well as its small size.
I would *love* to have a little wand I could poke into places I cannot see. I have a microscope, but it is big and bulky. The little USB microscope I have has serious depth of field issues, resulting in me being able to only see things "in focus" at any particular setting.
Now, if I could get something like a "test probe" that can see at the end, and let me wear one of those little displays on my glasses so I can see what my little probe is seeing, it would sure make a lot of my work a helluva lot easier.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2) by carguy on Saturday March 29 2014, @11:57AM
> Now, if I could get something like a "test probe" that can see at the end, ...
Hi tech replacement for an inspection mirror--sounds pretty useful. Mechanics and dentists use little mirrors, but there is always the problem of getting light on the subject. The next patent might be combining this camera with tiny white LEDs.
This could also vastly reduce the cost of (low-res?) IR cameras. My understanding is that a good part of the cost is in the fancy lens material that can pass IR (regular glass blocks longer wavelengths)--much smaller lens = lower cost.
(Score: 2) by umafuckitt on Saturday March 29 2014, @01:13PM
At what wavelengths does regular glass block IR? Is it far IR? I use regular, cheap, lenses at work for a 950 nm laser and there are no issues. My lenses are just coated for those wavelengths, but that's not expensive. Thor sell achromatic doublets rated to 1620 nm [thorlabs.com] for only about 45 bucks. Seems to be the same price [thorlabs.com] for visible wavelengths.
(Score: 1) by cosurgi on Saturday March 29 2014, @01:37PM
yeah, I concur. Actually the problem is on the other end of the spectrum. It is difficult to find glass that is transparent for wavelengths shorter than 300nm.
#
#\ @ ? [adom.de] Colonize Mars [kozicki.pl]
#
(Score: 1) by cafebabe on Sunday March 30 2014, @07:16PM
I believe Vernor Vinge described Fourier cameras somewhere in the compilation Across Realtime which was published in 1986.
1702845791×2