gabor is credited as being the one of the first to invent holography, leith for having innovated the use of a laser and later the off-axis technique.
Actually the off-axis technique came before laser holography, and the off-axis technique really is a big deal--a Nobel-Prize-worthy big deal--since the twin-image problem caused holography research to stagnate for years. By 1955, even Gabor had pretty much given up on holography because of the twin-image problem. After Leith and Upatnieks solved the twin-image problem by introducing the off-axis method in 1962, and after Leith and Upatnieks demonstrated the first three-dimensional hologram in 1964 (all of Gabor's holograms were two-dimensional), the field of holography exploded.
not making this distinction is like creditiing henry ford for the invention of the automobile.
Similarly there is a huge difference between saying you can do something and actually doing it. Gabor said it could be possible to make three-dimensional holograms, but no one (including Gabor) was actually able to make a three-dimensional hologram until Leith and Upatnieks made and displayed the very first one in 1964. You may have a looser definition of invention, but in my opinion it's not enough to say you can do something--you have to do it.
No, as the article says, Dennis Gabor invented holography and coined the term "hologram", in 1948.
My original statement is correct because of the words I have emphasized in bold face. Gabor only made holograms of two-dimensional objects (specifically of a transparency of the words "Hugyens", "Young", and "Fresnel" and of a transparent protractor). Admittedly part of the reason for this was because of the lack of a light source with the coherence length available from a laser, but part of the reason is because his method of recording holograms had a serious problem known as the "twin-image" problem. As a result, the development of holography stagnated until Prof. Leith invented a method for solving this problem. If you want to play semantic games regarding the exact meaning of the word "inventor", go ahead, but it is an indisputable fact that Leith and Upatnieks demonstrated the first three-dimensional hologram.
There's a reason Gabor won a Nobel prize.
I don't dispute this, but there is a less fortunate reason why Prof. Leith did not receive a share of the prize. A professor at Michigan who had a personal dislike for Prof. Leith (and who in fact tried repeatedly to steal credit for work that had actually been done by Prof. Leith) actively campaigned against giving Prof. Leith a share of the Nobel Prize. Despite this, Prof. Leith is actually mentioned by name in the speech awarding the Nobel Prize to Gabor, a highly-unusual (if not unique) occurrence in the history of the Nobel Prize.
In negative refraction, the refracted ray is on the "wrong" side of the surface normal. It looks like the light has been reflected by the surface normal, except there is no interface along the surface normal for the light to reflect off of.
I did a project investigating the performance of eigenface facial recognition for a graduate-level class in image processing last semester. I'll concede that eigenfaces are not the be-all end-all for face recognition (if they were, face recognition would be deployed everywhere), but they work pretty well as a first-order approach for tackling facial recognition and classification problems. It's not nearly as sensitive to scaling and orientation as you claim, and even if it were, you can combine it with feature detectors (to find the coordinates of the eyes and mouth, for example) to find and apply transformations to the images that reduce the effect of scaling and orientation.
As for facial expression being a poor metric of emotion, it's not that much worse than the alternatives. You can ask people, "How are you feeling?", but they can always lie in their responses. You could try to use vocal inflection, but that's probably a more difficult classification problem and probably a less reliable indicator. In this particular case, facial expression is probably the best available metric since we can't go out and ask Mona Lisa, "So, what emotion were you feeling when you posed for your portrait?"
I agree that face recognition is in its infancy, but to call it pseudo-scientific is crossing the line. The reality is that, in its full generality, face recognition is a very hard problem, so the best way to start tackling the general problem is to tackle simpler versions of the problem. The simpler versions of the problem may seem contrived, but solving the simpler versions of the problem may shed some light on how to attack more difficult versions of the problem. That is science in action.
The difference between what you describe and what the research group at Brown did is that, in your case, you get the same 630 nm light out. The Brown group used 514 nm light (probably from an argon laser) to drive the device and observed a luminescence signal with a center wavelength around 1278 nm, which is what one would expect for silicon.
I think your point is that the article is lacking details as to why the silicon needs to be nanopatterned in the first place, and even the preprint in Nature Materials fails to provide motivation for the nanopatterning. I suspect the nanopatterning either modifies the electronic bandgap in silicon (possibly making the nanopatterned silicon a direct bandgap material), or it creates a photonic bandgap that suppresses spontaneous emission or enhances stimulated emission, but that's just speculation on my part.
They reduce resolution by a factor of 180, but only improve depth-of-field by a factor 7. This is particularly silly because the only reason they have a bad depth-of-field is because they are using a huge expensive sensor. If they would switch to a small cheap sensor like you find in any cheap digicam (1/1.8"), they would get the same improvement, and save $14800.
Actually, depth-of-field is entirely controlled by the imaging optics. Using a larger or smaller array increases the field of view, but has no effect on the depth of field. One can improve the depth of field by stopping down the lens, but this also trades away lateral resolution (assuming the optics and sensor have equal resolution at full aperture, you're now wasting some of the resolution of your sensor), and it throws away light (which is never good for signal-to-noise ratio). Focal error also trades away lateral resolution and wastes the resolution of your sensor. Admittedly, this technique also trades away some of the resolution of the sensor for its refocusing capability (which can become extended depth of field with appropriate image processing), but it does so without throwing away light. If you're limited in what you can do with the optics, this technique lets you circumvent those limitations to some extent with the sensor and with image processing, so I think it's a bit hasty to say this technique can never find a practical application.
Stupid question - but what is a laser diode ? Granted it is doped silicon, but why should we care ?
Actually, most laser diodes are made of aluminum gallium arsenide (AlGaAs) or indium gallium arsenide phospide (InGaAsP). At the moment, there are no commercial silicon laser diodes because there are no silicon laser diodes. The reason, as is alluded to in the linked article, is that silicon is an indirect bandgap material, so a photon (a quantum of light or electromagnetic vibration) emission event can only occur if a phonon (a quantum of lattice vibration in the material--in this case, silicon) either arrives or leaves the location of the photon emission event at the same time. The probability of these events occuring simultaneously is vanishingly small, which makes optical processes in silicon extremely inefficient and makes it extremely difficult to use silicon as an optoelectronic material.
The UCLA laser was a Raman laser that could only operate in pulsed mode. The Raman effect is a nonlinear effect that requires several external laser beams to power the silicon device.
The Intel laser was also a Raman laser and was the first silicon Raman laser that could operate in continuous-wave (non-pulsed) mode.
The Brown laser is not a Raman laser. Therefore it only requires a single external laser beam to power the device.
The holy grail, of course, is an electrically-pumped silicon laser where you apply a voltage directly across the device and get laser light out. We're not there yet, but each of these devices represents progress toward that goal. In particular, a device with direct optical pumping like the Brown laser suggests that direct electrical pumping might not be far off.
Summary: Stupid Silicon Tricks candidate. No viable application.
Well, it's going to take a few clever silicon tricks to make optoelectronic devices out of silicon. The linked article alludes to silicon's indirect bandgap, which is why silicon is such a troublesome optoelectronic material. Progress in computing speed and communications bandwidth will stagnate until we either discover how to make optoelectronic devices out of silicon or we discover how to make logic gates out of InGaAsP or AlGaAs. Because all of the existing infrastructure for silicon, it would be preferrable to develop silicon as an optoelectronic material. Thus I consider this device to be a valuable first step towards a useful silicon laser.
Think about the barcode example. You don't need a priori knowledge at all because you know (1) the ray along which the laser is pointing and (2) the ray along which you have seen the point. It's fairly trivial to reconstruct the geometry.
You are implicitly making assumptions (and therefore asserting a priori knowledge) about the scene geometry. I agree that you know (1) the ray along which the laser is pointing, but in estimating (2) the ray along which you have seen the point, you are assuming that the scene is planar and a known distance and orientation relative to the source and detector. If that is the case, then I agree that the reconstruction is trivial. However if you do not allow those assumptions, then you only know where ray (2) intersects the detector plane, not where it intersects the scene, and therefore it becomes significantly more complicated to map your detected signal back to the scene.
It might be easier to understand the problem if you draw it out. The knowns are a line (representing ray (1)) and a point not on that line (representing the location of the detected signal, which is the intersection of ray(2) with the detector plane). If that's all you know, then you don't know the direction of ray (2), so you can't map the detected signal back to the scene. However, if you know where ray (1) intersected the scene (which requires a priori knowledge of the scene geometry), then I agree that you can easily determine ray (2).
Let me briefly address your point about ray tracing. In ray tracing, you know the scene geometry exactly (you have a complete description of it inside the computer!), so it is trivial render a new viewpoint of the scene. The Stanford group is doing this with no a priori knowledge of the scene, just the data that they collect. That is a pretty impressive feat.
One of the points that I was trying to make is that CT is not particularly complex to compute. Another is that there are many different imaging modalities that may be interesting or useful for particular applications. CT or DT are useful for three-dimensional transmission imaging of semitransparent three-dimensional objects. The Stanford group presents an interesting technique for imaging three-dimensional scenes in reflection, and estimating information that is not obviously present in ordinary photographs. Neither is "better" than the other in a general sense because they are designed for different applications.
Now compare this to a *really cool* imaging technique, like using an x-ray beam and an array of photodiodes to detect the scatter patterns as the beam passes through a human body, then calculate an image of the actual bones and organs inside. It's called Computed Axial Tomography or a CAT scan.
Actually, traditional computed tomography makes the assumption that x-rays do not scatter as they propagate through the specimen. Then, by the projection-slice theorem, reconstruction is fairly straightforward: Fourier transform the projections, orient them correctly in Fourier space, then inverse Fourier transform the total spectrum. The scattering case (known as diffraction tomography, and necessary when considering optical or ultrasonic tomography) is much more complicated, and generally requires making a weak scattering assumption like either the Born approximation or the Rytov approximation.
I can't RTFA, but I'm pretty sure that what you describe is not what they're doing. The remarkable claim that they make is that from images of a three-dimensional scene that are captured at a particular camera location, they can render an image that the camera would have seen from a different location (namely the location of the illuminator). Furthermore, they do this without a priori knowledge of the scene geometry. In your barcode example, you need a priori knowledge of the position of the source and the camera to correctly re-render the image from the perspective of the source.
What exactly is going to wear-and-tear in a flash drive? Since they are solid-state storage devices, and therefore have no moving parts, most failures occur early and failure rate actually decreases with elapsed time.
It took me all of 60 seconds to Google this link subtantiating the factor of 3-5 slowdown with Mudflap: http://gcc.fyxm.net/summit/2003/mudflap.pdf The performance data is tabulated on page 7: the average slowdown out of six test cases (three build case, three run cases) appears to be a factor of 4 or so, with the best case being 1.25 (in one run case), and the worst case being 5 (in one build case and in one run case).
why dont you shine a projector towards a mirror and have a look what you see in the mirror, thats right! a projector!
Shine a projector onto a plane of specularly-reflecting material (like a mirror) and you see the projector. Shine a projector onto a plane of diffusely-reflecting material (like a white-painted wall or a projection screen) and you see the projected image.
It could make sense if the primary component is a holographically-recorded optical element. However I'm reserving the term "holographic video" for this display,whenever it becomes commercially available.
I think I've worked out the basic details of how this screen works from this link. The heart of the screen is a diffractive optical element, holographically-recorded on a thin photopolymer layer. Based on the range of acceptance angles, I think the element is the hologram of a 27-degree deviating prism. The viewing angle problem is solved by placing a weak diffuser (an example of a strong diffuser is frosted glass) on the viewing side, either in contact with or in close proximity to the diffractive optical element. The diffuser scatters the projected light over some relatively narrow range of angles (about 25 degrees, according to this link). The close proximity of the diffuser also takes care of the dispersion problem because it doesn't give the different colors much distance over which to spread out, and that spread is masked by the angular spread introduced by the diffuser.
From the article (which is rather lacking in technical details), the display sounds like a holographically-recorded diffractive optical element on a glass substrate. If so, I'm curious how they compensate for the dispersion intrinsic to the diffraction phenomenon (since selling a 15,000 quid monochrome display is probably not a commercially-viable option:-p). Also, since the display claims to be angularly selective (it has to be if it only accepts a specific projection direction), I wonder if it has a similarly selective viewing angle (like early LCD displays, which were only bright and clear at normal incidence).
I rebut that link in this reply to one of your other comments. The feedback due to the mirrors does contribute to the spatial coherence of the beam, but ultimately it depends on the fact that the stimulated emission process is temporally-coherent (in-phase). A Q-switched laser is a simple counterexample (at least to an optical physicist). Another simple counterexample that I neglected to mention in that other post is this. According to your theory, if I remove the laser medium from the laser and shine light into the former laser from one end, the output from the other end should be spatially-coherent light. This non-laser laser is actually a device called a Fabry-Perot interferometer, and it does not cause spatially-incoherent light to become spatially-coherent (also a simple experiment to do in an optics lab). Thus feedback is insufficient to explain the spatial coherence of a laser.
For a proper treatment of coherence, I recommend Statistical Optics by Joseph Goodman, or if you're a masochist you can attempt to tackle Optical Coherence and Quantum Optics by Leonard Mandel and Emil Wolf.:-p
Slashdot Mirror
In negative refraction, the refracted ray is on the "wrong" side of the surface normal. It looks like the light has been reflected by the surface normal, except there is no interface along the surface normal for the light to reflect off of.
I did a project investigating the performance of eigenface facial recognition for a graduate-level class in image processing last semester. I'll concede that eigenfaces are not the be-all end-all for face recognition (if they were, face recognition would be deployed everywhere), but they work pretty well as a first-order approach for tackling facial recognition and classification problems. It's not nearly as sensitive to scaling and orientation as you claim, and even if it were, you can combine it with feature detectors (to find the coordinates of the eyes and mouth, for example) to find and apply transformations to the images that reduce the effect of scaling and orientation.
As for facial expression being a poor metric of emotion, it's not that much worse than the alternatives. You can ask people, "How are you feeling?", but they can always lie in their responses. You could try to use vocal inflection, but that's probably a more difficult classification problem and probably a less reliable indicator. In this particular case, facial expression is probably the best available metric since we can't go out and ask Mona Lisa, "So, what emotion were you feeling when you posed for your portrait?"
I agree that face recognition is in its infancy, but to call it pseudo-scientific is crossing the line. The reality is that, in its full generality, face recognition is a very hard problem, so the best way to start tackling the general problem is to tackle simpler versions of the problem. The simpler versions of the problem may seem contrived, but solving the simpler versions of the problem may shed some light on how to attack more difficult versions of the problem. That is science in action.
The difference between what you describe and what the research group at Brown did is that, in your case, you get the same 630 nm light out. The Brown group used 514 nm light (probably from an argon laser) to drive the device and observed a luminescence signal with a center wavelength around 1278 nm, which is what one would expect for silicon.
I think your point is that the article is lacking details as to why the silicon needs to be nanopatterned in the first place, and even the preprint in Nature Materials fails to provide motivation for the nanopatterning. I suspect the nanopatterning either modifies the electronic bandgap in silicon (possibly making the nanopatterned silicon a direct bandgap material), or it creates a photonic bandgap that suppresses spontaneous emission or enhances stimulated emission, but that's just speculation on my part.
The holy grail, of course, is an electrically-pumped silicon laser where you apply a voltage directly across the device and get laser light out. We're not there yet, but each of these devices represents progress toward that goal. In particular, a device with direct optical pumping like the Brown laser suggests that direct electrical pumping might not be far off.
Grilling meat produces chemical compounds called heterocyclic amines which are known carcinogens.
Likewise, I will concede that CT and MRI are pretty cool. I just happen to think this is pretty cool too. ;-)
It might be easier to understand the problem if you draw it out. The knowns are a line (representing ray (1)) and a point not on that line (representing the location of the detected signal, which is the intersection of ray(2) with the detector plane). If that's all you know, then you don't know the direction of ray (2), so you can't map the detected signal back to the scene. However, if you know where ray (1) intersected the scene (which requires a priori knowledge of the scene geometry), then I agree that you can easily determine ray (2).
Let me briefly address your point about ray tracing. In ray tracing, you know the scene geometry exactly (you have a complete description of it inside the computer!), so it is trivial render a new viewpoint of the scene. The Stanford group is doing this with no a priori knowledge of the scene, just the data that they collect. That is a pretty impressive feat.
One of the points that I was trying to make is that CT is not particularly complex to compute. Another is that there are many different imaging modalities that may be interesting or useful for particular applications. CT or DT are useful for three-dimensional transmission imaging of semitransparent three-dimensional objects. The Stanford group presents an interesting technique for imaging three-dimensional scenes in reflection, and estimating information that is not obviously present in ordinary photographs. Neither is "better" than the other in a general sense because they are designed for different applications.
I can't RTFA, but I'm pretty sure that what you describe is not what they're doing. The remarkable claim that they make is that from images of a three-dimensional scene that are captured at a particular camera location, they can render an image that the camera would have seen from a different location (namely the location of the illuminator). Furthermore, they do this without a priori knowledge of the scene geometry. In your barcode example, you need a priori knowledge of the position of the source and the camera to correctly re-render the image from the perspective of the source.
What exactly is going to wear-and-tear in a flash drive? Since they are solid-state storage devices, and therefore have no moving parts, most failures occur early and failure rate actually decreases with elapsed time.
It took me all of 60 seconds to Google this link subtantiating the factor of 3-5 slowdown with Mudflap: http://gcc.fyxm.net/summit/2003/mudflap.pdf The performance data is tabulated on page 7: the average slowdown out of six test cases (three build case, three run cases) appears to be a factor of 4 or so, with the best case being 1.25 (in one run case), and the worst case being 5 (in one build case and in one run case).
"Reify"...I haven't seen that word since I briefed Critical Legal Studies for high-school debate... :-D
I think I've worked out the basic details of how this screen works from this link. The heart of the screen is a diffractive optical element, holographically-recorded on a thin photopolymer layer. Based on the range of acceptance angles, I think the element is the hologram of a 27-degree deviating prism. The viewing angle problem is solved by placing a weak diffuser (an example of a strong diffuser is frosted glass) on the viewing side, either in contact with or in close proximity to the diffractive optical element. The diffuser scatters the projected light over some relatively narrow range of angles (about 25 degrees, according to this link). The close proximity of the diffuser also takes care of the dispersion problem because it doesn't give the different colors much distance over which to spread out, and that spread is masked by the angular spread introduced by the diffuser.
From the article (which is rather lacking in technical details), the display sounds like a holographically-recorded diffractive optical element on a glass substrate. If so, I'm curious how they compensate for the dispersion intrinsic to the diffraction phenomenon (since selling a 15,000 quid monochrome display is probably not a commercially-viable option :-p). Also, since the display claims to be angularly selective (it has to be if it only accepts a specific projection direction), I wonder if it has a similarly selective viewing angle (like early LCD displays, which were only bright and clear at normal incidence).
I rebut that link in this reply to one of your other comments. The feedback due to the mirrors does contribute to the spatial coherence of the beam, but ultimately it depends on the fact that the stimulated emission process is temporally-coherent (in-phase). A Q-switched laser is a simple counterexample (at least to an optical physicist). Another simple counterexample that I neglected to mention in that other post is this. According to your theory, if I remove the laser medium from the laser and shine light into the former laser from one end, the output from the other end should be spatially-coherent light. This non-laser laser is actually a device called a Fabry-Perot interferometer, and it does not cause spatially-incoherent light to become spatially-coherent (also a simple experiment to do in an optics lab). Thus feedback is insufficient to explain the spatial coherence of a laser.
:-p
For a proper treatment of coherence, I recommend Statistical Optics by Joseph Goodman, or if you're a masochist you can attempt to tackle Optical Coherence and Quantum Optics by Leonard Mandel and Emil Wolf.