Re:The holy grail of camera tech....
on
HDR Video a Reality
·
· Score: 3, Interesting
You don't need aperture at all if you use a microlens array to do integral photography. On top, you get full depth (3D) and capture all focal lengths, including the focal depth information. All in a single shot. Just need an ultra high resolution sensor--or, instead, an array of many small cameras (works just as well, and no need for perfect alignment as that can be finessed in software). You capture a full 4D lightfield (light can be parameterized as the two pairs of coordinates of a light ray crossing two infinite planes), i.e. miss no optical information whatsoever other than your diffraction and wavelength limits.
Re:The holy grail of camera tech....
on
HDR Video a Reality
·
· Score: 4, Interesting
You forgot about full lightfield capture. This can be done with a single camera using ultra high resolution and a microlens array (or alternatively, an array of a very large number of tiny cameras). Think single camera, single shot capture of depth (3D) and all focus planes. Then you can reproduce the full 3D and multiple focus depths (as in, the eye would have to focus at different depths) on a flat display with microlens array covering it (again, need ultra-high resolution since focal depths and parallax viewpoints are discretized to the pixel number covered by each micro lens).
In fact, hidden content can be interpolated and is done with texture seam filling algorithms, sufficient for a limited motion parallax. Moreover, there are two other ways motion parallax can be dealt with: using more cameras, and even simply using a wider separation with just the usual two cameras (the content then can be reprojected for a smaller separation based on the derived depth-map).
Mircolens can in fact handle adaptation, so you are 100% wrong on that: http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1545799. Also note that microlens handle not just accommodation but naturally reproduce motion parallax. This is all accomplished with a single camera which also has a microlens array (or just use an array of small cameras of standard resolution). It's unfortunate I did not see your post earlier so I could correct the horrible misinformation on microlens you perpetrated upon the/. reading public.
You seem to be not aware of the huge difference correct 3D positional audio can give. If you have good headphones, try listening to a binaural recording. The 3D reproduction is on a whole different level from Dolby etc.
Uh, dude, 3D without glasses using as standard tech as LCD displays has been around for over a decade. Lenticular arrays and parallax barrier are very old tech by now.
Re:The "sweet spot" problem and the "edge" problem
on
The Joke Known As 3D TV
·
· Score: 2, Interesting
In fact, multichannel audio recordings (and I include stereo into that) have the exact same sweet spot problem, because even with the best recording practices trying to capture 3D (i.e. Ambisonics which encodes a 3D soundscape with spherical harmonics) can only have _correct_ reproduction in a small sweet spot. And yet, multichannel audio is the standard, because it contributes something even if that something is very far from the optimum.
Objects can never go outside the screen of any display system, because light movies in straight lines, and thus there has to be a display element behind every pixel you see displayed. This is the case with anything from TV to holographic displays. Unless you add some black holes whose gravity make the light go in curved paths, a part of the display will always be behind every piece of visual information.
Disparity is the most important factor in 3D perception in the human visual system, other than motion parallax. Two cameras capture disparity, and motion parallax can be achieved with head-tracking--technology to do this with computer vision instead of having to use a head-mounted display has existed for the past 20 years at least. Vergence and accommodation (focus) are secondary and always overridden by the other factors; this is a neurophysiological fact. Moreover, both can be handled by the use of an ultra-high resolution display and a microlens array--or simply using a head-mounted display with active optics and eye tracking. And the information from two cameras is sufficient from both, because depth can be extracted from disparity and all other effects can be computed for a fitting display.
While Brightside's showcase product at SIGGRAPH etc. was a large screen TV, the original prototype addressable LED backlit HDR display (developed in the comp. sci. graphics dept. at UBC so I got to play around with it) was a computer monitor. As for blooming, it turned out to be a minor issue because, while the human eye has a very high dynamic range overall, locally over a small arc of the visual field its dynamic range is much lower. This way the LED backlight array could simply be powered by a low-passed version of the image.
Of course, you're both playing semantics games. In a von Neumann machine, such as is every desktop computer, for example, the separation of data and program is superficial--it's just a psychologically-driven convention. It is also an extremely frequently violated convention (both by machine--Windows tends to rewrite memory-loaded images of binaries heavily--and by humans, in cases not just of the more rare virus-modifying code, but in every instance of scripting/interpreters/just-in-time compilation). Thus, obscuring the keys is not fundamentally different from obscuring the algorithms because there is no fundamental distinction between program and data. In practical terms it may be more convenient to have many keys per algorithm rather than the other way around, but this is merely adopted for trivial practical reasons. Again, there's nothing wrong with "through obscurity" by the usual definition as long as the level of obscurity applied to the algorithms corresponds to the level of obscurity applied to the keys in the more common approach.
The computable reals is a countable set, whereas reals are not, and thus essentially all reals are not computable. Note that computable reals are necessary to have a true Turing machine, but the problem is that quantum uncertainty will break any infinite time algorithm--the probability that your computing system fails goes to 1 as time goes to infinity. So we are still stuck with linearly bounded automata being the most powerful information processing machines (or brains, for that matter) that can be constructed in practice.
Fail! You forget the holographic principle and Bekenstein bound, which are largely accepted in physics today. You can only have a finite number of distinguishable quantum states in a bounded region of space. It directly follows that real numbers cannot have a direct equivalent in the physical universe because they allow a violation of this bound, since a real number allows you to encode an infinite amount of information (most real numbers require an infinite number of digits to express, it's infinite precision). Another thing that follows is that even in an infinite spacetime extent universe, there can only be countable infinities. It follows further that any machine in the universe is at best a linearly bounded automata (not even a full Turing machine, since any system is spacelike bounded by its light cone, which does not in practice arbitrarily expand in the future, since as time approaches infinity, quantum uncertainty guarantees that the probability that a critical failure will occur in the system approaches unity).
Since light only travels in straight lines, it is impossible to see an image from even a holographic display outside the visual boundaries of the projection system. There must always be a laser/monitor/spinning mirror/fog/whatever behind EVERY SINGLE PIXEL of the holographic image. Only intrusive display systems can get around that--direct retinal projection (laser 'painting' an image on your retina). or electrical stimulation of the optic nerve through the retina.
Please mod the parent down, as the post implies that people using autostereoscopic displays are safer, and this could be a hazard to their vision. It is the difference between stereopsis (convergence) and accommodation (focus) that is the issue, and except two types of 3D will suffer from this: holographic and volumetric (and possibly specially configured microlensarray displays)
OpenGL 3 was finished a long time ago, and is now up to version 3.3. For the newer hardware, 4.0 was finished some months ago as a separate branch. Drivers for both have been out from NVIDIA as soon as the new specs were released.
Why should it be more widely used, when it's as flawed as other voting systems? It fails monotonicity and independence of irrelevant alternatives. Who are you to say that these voting criteria are less important than others that STV satisfies but some other voting systems do not? In the end, the choice of which voting criteria to compromise on is a purely subjective choice.
Slashdot Mirror
You don't need aperture at all if you use a microlens array to do integral photography. On top, you get full depth (3D) and capture all focal lengths, including the focal depth information. All in a single shot. Just need an ultra high resolution sensor--or, instead, an array of many small cameras (works just as well, and no need for perfect alignment as that can be finessed in software). You capture a full 4D lightfield (light can be parameterized as the two pairs of coordinates of a light ray crossing two infinite planes), i.e. miss no optical information whatsoever other than your diffraction and wavelength limits.
You forgot about full lightfield capture. This can be done with a single camera using ultra high resolution and a microlens array (or alternatively, an array of a very large number of tiny cameras). Think single camera, single shot capture of depth (3D) and all focus planes. Then you can reproduce the full 3D and multiple focus depths (as in, the eye would have to focus at different depths) on a flat display with microlens array covering it (again, need ultra-high resolution since focal depths and parallax viewpoints are discretized to the pixel number covered by each micro lens).
In fact, hidden content can be interpolated and is done with texture seam filling algorithms, sufficient for a limited motion parallax. Moreover, there are two other ways motion parallax can be dealt with: using more cameras, and even simply using a wider separation with just the usual two cameras (the content then can be reprojected for a smaller separation based on the derived depth-map).
/. reading public.
Mircolens can in fact handle adaptation, so you are 100% wrong on that: http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1545799. Also note that microlens handle not just accommodation but naturally reproduce motion parallax. This is all accomplished with a single camera which also has a microlens array (or just use an array of small cameras of standard resolution). It's unfortunate I did not see your post earlier so I could correct the horrible misinformation on microlens you perpetrated upon the
You seem to be not aware of the huge difference correct 3D positional audio can give. If you have good headphones, try listening to a binaural recording. The 3D reproduction is on a whole different level from Dolby etc.
Uh, dude, 3D without glasses using as standard tech as LCD displays has been around for over a decade. Lenticular arrays and parallax barrier are very old tech by now.
In fact, multichannel audio recordings (and I include stereo into that) have the exact same sweet spot problem, because even with the best recording practices trying to capture 3D (i.e. Ambisonics which encodes a 3D soundscape with spherical harmonics) can only have _correct_ reproduction in a small sweet spot. And yet, multichannel audio is the standard, because it contributes something even if that something is very far from the optimum.
Objects can never go outside the screen of any display system, because light movies in straight lines, and thus there has to be a display element behind every pixel you see displayed. This is the case with anything from TV to holographic displays. Unless you add some black holes whose gravity make the light go in curved paths, a part of the display will always be behind every piece of visual information.
Disparity is the most important factor in 3D perception in the human visual system, other than motion parallax. Two cameras capture disparity, and motion parallax can be achieved with head-tracking--technology to do this with computer vision instead of having to use a head-mounted display has existed for the past 20 years at least. Vergence and accommodation (focus) are secondary and always overridden by the other factors; this is a neurophysiological fact. Moreover, both can be handled by the use of an ultra-high resolution display and a microlens array--or simply using a head-mounted display with active optics and eye tracking. And the information from two cameras is sufficient from both, because depth can be extracted from disparity and all other effects can be computed for a fitting display.
A paranoid conspiracy theorist on Slashdot? Who da thunk it!
Please mod parent down: as of 2009, 80% of blackberry customers are non-corporate consumers: http://www.twice.com/article/295368-RIM_Majority_Of_BlackBerry_Users_Now_Consumers_Small_Businesses.php
Looks like /. formatted the period out of the link I posted. The correct URL includes a period at the end of 'Inc.'--http://en.wikipedia.org/wiki/BrightSide_Technologies_Inc.
Parent wrote: "As far as I know, there isn't a single computer monitor that allows local area dimming"
Shows what you know. http://en.wikipedia.org/wiki/BrightSide_Technologies_Inc.
While Brightside's showcase product at SIGGRAPH etc. was a large screen TV, the original prototype addressable LED backlit HDR display (developed in the comp. sci. graphics dept. at UBC so I got to play around with it) was a computer monitor.
As for blooming, it turned out to be a minor issue because, while the human eye has a very high dynamic range overall, locally over a small arc of the visual field its dynamic range is much lower. This way the LED backlight array could simply be powered by a low-passed version of the image.
self-modifying **
Of course, you're both playing semantics games. In a von Neumann machine, such as is every desktop computer, for example, the separation of data and program is superficial--it's just a psychologically-driven convention. It is also an extremely frequently violated convention (both by machine--Windows tends to rewrite memory-loaded images of binaries heavily--and by humans, in cases not just of the more rare virus-modifying code, but in every instance of scripting/interpreters/just-in-time compilation). Thus, obscuring the keys is not fundamentally different from obscuring the algorithms because there is no fundamental distinction between program and data. In practical terms it may be more convenient to have many keys per algorithm rather than the other way around, but this is merely adopted for trivial practical reasons. Again, there's nothing wrong with "through obscurity" by the usual definition as long as the level of obscurity applied to the algorithms corresponds to the level of obscurity applied to the keys in the more common approach.
Ditto.
The computable reals is a countable set, whereas reals are not, and thus essentially all reals are not computable. Note that computable reals are necessary to have a true Turing machine, but the problem is that quantum uncertainty will break any infinite time algorithm--the probability that your computing system fails goes to 1 as time goes to infinity. So we are still stuck with linearly bounded automata being the most powerful information processing machines (or brains, for that matter) that can be constructed in practice.
There is no amount of physics you can do to need to worry about Godel--see my reply to the parent.
Fail! You forget the holographic principle and Bekenstein bound, which are largely accepted in physics today. You can only have a finite number of distinguishable quantum states in a bounded region of space. It directly follows that real numbers cannot have a direct equivalent in the physical universe because they allow a violation of this bound, since a real number allows you to encode an infinite amount of information (most real numbers require an infinite number of digits to express, it's infinite precision). Another thing that follows is that even in an infinite spacetime extent universe, there can only be countable infinities. It follows further that any machine in the universe is at best a linearly bounded automata (not even a full Turing machine, since any system is spacelike bounded by its light cone, which does not in practice arbitrarily expand in the future, since as time approaches infinity, quantum uncertainty guarantees that the probability that a critical failure will occur in the system approaches unity).
Since light only travels in straight lines, it is impossible to see an image from even a holographic display outside the visual boundaries of the projection system. There must always be a laser/monitor/spinning mirror/fog/whatever behind EVERY SINGLE PIXEL of the holographic image. Only intrusive display systems can get around that--direct retinal projection (laser 'painting' an image on your retina). or electrical stimulation of the optic nerve through the retina.
Ooops for not proofreading... the last sentence clause should have been "and ALL except two types of 3D DISPLAYS will..."
Please mod the parent down, as the post implies that people using autostereoscopic displays are safer, and this could be a hazard to their vision. It is the difference between stereopsis (convergence) and accommodation (focus) that is the issue, and except two types of 3D will suffer from this: holographic and volumetric (and possibly specially configured microlensarray displays)
OpenGL 3 was finished a long time ago, and is now up to version 3.3. For the newer hardware, 4.0 was finished some months ago as a separate branch. Drivers for both have been out from NVIDIA as soon as the new specs were released.
Fail. This is still subject to Arrow's impossibility theorem.
Why should it be more widely used, when it's as flawed as other voting systems? It fails monotonicity and independence of irrelevant alternatives. Who are you to say that these voting criteria are less important than others that STV satisfies but some other voting systems do not? In the end, the choice of which voting criteria to compromise on is a purely subjective choice.
This voting system is just as flawed as FPTP: as another post already mentioned--
http://en.wikipedia.org/wiki/Cumulative_voting#Voting_systems_criteria
http://en.wikipedia.org/wiki/Cumulative_voting#Tactical_voting