h a l f b a k e r yCaution!
Contents may be not!
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Five ideas, some maybe impractical, separable, labelled.
(i) Getting polarisation data
(ii) Hue
(iii) Storage
(iv) Manipulation
(v) Display
Digital images omit polarisation data, colour spaces distort short wavelengths, and cameras filter out IR and UV. Displays and digital
cameras ignore polarisation. It'd be nice to keep more information, display it and store it in a new way.
(i) Polarisation. Leave shutters, if they still exist, open longer and rotate a polarising or liquid crystal filter 'twixt lens and CCD mechanically or electrically. Time rotation, record receptor response, divide by a hundred and eighty, and store pixel response nature and time.
(ii) Hue. Rather than filter out IR or UV, record where the receptors respond to the whole wavelength gamut. Record it as modified HSV, storing hue as a number between zero and three hundred and fifty-nine. Merge short and long wavelengths merge at ultraviolet/infrared, i.e. invisible, rather than purple, leaving all visible hues as truly spectral, avoiding a magenta catastrophe but keeping near IR and UV data. IR, UV and true indigo and violet compensate for loss of colour depth. Use IR and UV filters to extract short and long wavelengths.
This provides data on intensity and a wider wavelength range as HSV with separate polarisation angle data. There'd be up to a hundred and eighty values per pixel, though often equal. I know little of data compression, but methinks rather than storing everything, it can be compressed to remove redundancy.
(iii) Store the rest as values between zero and a hundred and eighty in binary coded d o z e n a l - each dozenal digit a nybble with two bits left - more complex polarisation data augment ten-bit pixels, making thirty bits for HSV, i.e. nearly four bytes, less than RGB plus transparency. However, i think this is uncompressible by standard algorithms which are for human vision. Don't know about TIFF though. Each file is much bigger anyway. On the plus side, since they're stored as binary-coded dozenal, manipulation may be easier and faster.
(iv) These data can be interestingly processed, even just on the polarisation. Chirality can be detected and displayed, reflections and glare can be removed, skies can be bluer, specular and diffuse reflection can be distinguished, stresses in transparent materials are discernable and light sources outside the field of view can be reconstructed. All that useful information is now needlessly thrown away.
(v) Rear projection display. Align LEDs and electrically rotated liquid crystal pixels. The image looks normal but has another use. Take two unpolarised images simultaneously from slightly different angles and slice them into columns of pixels. Add polarisation data to each stripe at right angles, stick them back together and throw away half the stripes. Display the result and look at it through polarised glasses for a true-colour stereoscopic display. The data could be CGI.
messages in the wallpaper
http://en.wikipedia.../wiki/Steganography
[normzone, Apr 21 2009]
Polarization Imaging
http://www.opticsin...fm?URI=ao-20-9-1537
A review of the state of the art in 1981 [csea, Apr 21 2009]
DOLPi - a DIY polarimetric camera
https://www.google....polarimetric+camera
Implements (i). Older version used mechanically switched polarizers at different angles; current version uses a liquid crystal light valve as an electronically controllable polarization filter [notexactly, May 28 2019]
Mantis Shrimp (stomatopod) Eyes
https://www.nature....rticles/ncomms12140
"...Particular to each of these three sections are intra-retinal adaptations to the basic photoreceptor anatomy, which have enabled stomatopods to evolve regional specializations for both linear and circular polarization vision, as well as 12-channel colour vision..." [zen_tom, May 28 2019]
[link]
|
| |
"Hey, check out my new Canon Umptyscrunch! It's got 32 gigs of storage!"
"Awesome! How many pictures does it hold?"
"...one. But it's a really, really good one." |
|
| |
I like what I can make sense of in a quick read, especially the bit about extraction of polarization info for stress measurement. + |
|
| |
Would it be possible to remove the redundancy in the idea, (seems it has been pasted twice) or is it there for stereo imaging? ;) |
|
| |
so the duplicate text is accidental? the more interesting idea buried in the text is the polarized stereoscopic display. |
|
| |
Whoops, sorry about that. Still, ironic, isn't it?
I did think about posting this as several separate ideas. I will try to abbreviate it as well.
[phoenix], it wouldn't be quite that enormous. For example, a reflection on a window would be two images, and i suspect the situation is usually either completely unpolarised light, light polarised at a specific angle or light at two angles. I don't think anything like JPEG would work for this because it groups pixels together and throws away too much information. |
|
| |
Sorry [nineteenthly] - too lazy to read it all! BUT giving you this croissant because subtitle suggests that it's a good idea. |
|
| |
Thanks, [xenzag], it's a bit shorter now. It may even become the incredible shrinking idea.
[Normzone], you remind me of a place i used to live which really did have a polarised message on the wallpaper. Just outside the bathroom, there was a faint outline of the words "DEATH IS NEAR" only visible by diffuse reflection. When a housemate pointed it out, it seriously freaked me out. I still don't know how it got there but i wonder if it had a subliminal effect on whomever saw it. |
|
| |
There are cameras that use rotating color filters to capture color information instead of having a mosaique of differently filtered pixels on the chip. Your idea would call for even more filters (either on the rotating filter or on the chip, making either spatial or temporal resolution much worse - or both, if you combine the techniques. But i like the idea of total information overflow - Make a filter with every whole wavelength from 200nm to 1200nm, rotate that step by step, have a 180-polarisations (-90 and +90 are the same, so no need for 360) filter in front of that, for every wavelength rotate the pol-filter through the whole 180°. So 180.000 pol/col combinations per picture, not including circular polarisations. |
|
| |
Biological colour receptors respond to specific wavelengths and the vision of some animals can detect polarisation. Though i'm talking about filters here, they may not be necessary. I would expect the main issue to be temporal rather than spatial. I like your point about the halving of the angle, since that saves a bit per pixel as well as halving the data right away. It also occurs to me that in a way i'm talking about layers. Rather than storing each pixel in lots of different versions, the image could be compressed by simply recording the differences between differently polarised images and tagging them with a number indicating the angle. For instance, in an image of a lake sparkling in the sunlight under a blue sky, one layer could store the lake without the specular reflection, another the sky without the glare, a third the specular reflection alone and a forth the glare of the sky, assuming the angles of polarisation to differ.
So, you like the extra frequencies? They could also store quite significant data. For instance, an image of a naked person would show their superficial blood vessels in near infrared and in some cases subclinical dermatological lesions and sometimes even subcutaneous undetected tumours. Pretty important information, really. |
|
| |
(+)
I totally wasted twenty minutes picking out differences between the first copy and the second and then another fifteen or so trying to decipher an anagram of;
an piece which is I shall.
<heavy sigh> |
|
| |
I now have an ambition to make it the shortest idea on the HB, but that's not going to happen.
The Rearrangement Servant says "A A chillies which snipe." |
|
| |
Doh!
I didn't know that there was an anagram generator.
<puts check mark in Todays New Thing box and heads off to work> |
|
| |
Yes, though it seems to do it in alphabetical order, which is a bit irritating. |
|
| |
A good way to realize this might be "polarization bracketing" instead of exposure bracketing with a polarized filter. (i.e., you have 180 (or fewer) fast exposures with the polarization filter rotating) |
|
| |
Yes, good idea. The polarisation filter could be motorised and with a stationary subject this would allow it to be done with an unmodified camera. |
|
| |
[+] But why measure only linear polarization, but not circular or elliptical polarization? |
|
| |
I stand by my similar protestations on similar ideas. If you measure polarisation, you are not measuring polarisation, only the polarisation which you measure. This will make about as much sense as most of my other annos, but it is the truth. |
|
| |
As to the storing of "the whole gamut" of frequencies. This does already happen. Most notably in our orbiting satelite telescopes, certain medical equipment, and high end imaging mechanisms. The data is seldom stored in any compressed or aggregated sense. Software does this on the entire data set. Keeping original data sets is very important. If not just for what we can do with it now, but also for what we may be able to do with it later. data storage is now a commodity in which supply (at least not commercial supply) will never be above demand. |
|
| |
I have no doubt that our technology will take us to a point where we will lament the absence of polarisation RAW image data in current and legacy images. We are just not at the point we we can a) capture it *accurately*, b) disseminate it fully. You must bear in mind there is a whole arena of physicists that believe this may never be possible. |
|
| |
As to the "dozenal" data storage mechanism, I have long been working within the area of simplest representation of information (we call it the chocolate machine). I can tell you binary is it! Duodecimal/dozenal, hexadecimal, decimal, all will be recreated most simply within binary in terms of use. It has actually been proven, with a certain caveat. However the caveat only applies to unary (base one, no information), or base infinity (as long as it is countable), neither of which have any use. |
|
| |
//Chirality can be detected// Not so, mon frere. In the detection of a chirality you can change it. Not to say you always do, but the probability is that you do (pretty much every time). So in this case all you will be detecting is the change *you* made in the detection of what you were trying to detect. And then storing that in an over complicated system that will lead to data loss. |
|
| |
The idea, however, is noble. And completely halfbaked... |
|
| |
Thanks, [4whom]. Don't let the following rant detract from your accurate impression of my approval of your positive judgment of my idea. |
|
| |
When i said chirality, i was talking about stereochemistry, not the likes of helicity. I mean the likes of pointing a camera at syrup and determining its concentration, not weak decay doobries. I know little, but i would imagine something would stop that sort of thing from being sensible to measurement. Then again, i have a sneaking respect for hylomorphism which is giving me some cause for concern in my current attempts to help children learn chemistry, but that way madness lies, along with elephants being split in half by mahogany trees. Not gonna go there here.
Concerning duodecimal (i used "dozenal" to shorten the character count), i'm sure binary's efficient if what you're doing with it is straightforward binary arithmetic, but could you provide me with some clarification? Are you talking about efficiency with respect to storage space alone or in terms of processing speed? If the former, clearly that is so. If the latter, however, are you actually saying it can be proven that binary arithmetic is faster than looking duodecimal sums and products up in a table somewhere? I suspect that internally, colour data are stored as RGB. If you wanted to do something like rotate a hue through a right angle on a colour wheel, doing that with RGB feels like it'd be somewhat cumbersome. If, on the other hand, it's actually stored as HSV or some variant thereof, rotating it through a right angle is adding or subtracting seven and a half dozen from it. That would turn, for example, red to blue, which in RGB is presumably a no-brainer, but also lime-green to orange. Is there really an obvious method of doing both of those at the same time with RGB?
Whereas high-end imaging may capture a wider range of frequencies now, the CCDs in cheap camera 'phones can also do it now. It's just that the insensible light is filtered out. What i don't know, and probably should, is whether it would mess up the picture as it stands and if it can be disentangled. However, biological vision can do a lot of that already, so why not digital systems? |
|
| |
On Chiraltiy: I knew I made a mistake as I uttered the words. However they related to your words. |
|
| |
Photons are the most "unchiral" of particles. Heck they are even their own anti-particles. Polarisation is what we are really talking about. Light has *all* the kinds of polarisation, in *all* planar directions. There are those that think that these polarisations happen in more than a 2d plane (eg n-plane) as noted by [goldbb]. Let's say you want to measure this 2d-planar polarisation. Currently, and probably forever more, the only way can measure it is by observing it. The nature of observation of photons by photons changes what you are looking at. You can do this the old way with some kind of filter, but then you interact before measurement. The latest optical systems use a known polarity (laser) to introduce a known "unknown" and then deduce the result. Disturbingly they are only right a certain probability (currently approaching 50%) of the time. That actually seems to be the theoretical limit, hence quantum optical cryptography. |
|
| |
Well, in a couple of weeks time i'm going to try something with D- and L-limonene which should, so to speak, shed some light on that, and as it happens i was wondering about what would happen if i used a laser. |
|
| |
Onto Binary representation: |
|
| |
Let's start simple. You want to count all numbers consecutively. You have a system of light switches with which you want to store this data. Light switches have two positions, namely *on* and *off*. You start with decimal because that is how you currently count. You soon notice that you need a ten state switch (or a few ten state switches) to count like that. |
|
| |
You now have a decision to make. Up the complexity of the switch (ergo up the complexity of your measurements) , or reduce the complexity (i.e base) of your counting. It doesn't really matter which one you choose, the result will be the same. |
|
| |
This has to be viewed with some constraints imposed. Or else base 10, or higher and beyond, seems like a good choice. Unfortunatley the physical mechanism of measuring larger state systems operates on something larger than the square, and more often larger than the cube of the possible states. Until we discover, and accurately measure the n-planes of photon polarisation, we are not going to have more than binary states. Of course we could have unary states, but this imparts no information, whatever goes in unary stays in in the unary (hawking radiation excluded in this sense.) |
|
| |
It has long been a drawing card as to how biological entities can accept, parse/assimilate so much information given a baud rate of (i think) 9000 odd bps. It is just that we , bio processors, receive analog (base infinity) information. |
|
| |
I am not saying it is impossible for more information to be stored per unit "information carrying unit" than binary. I am just saying it is impossible to generate less than a binary representation of the data in a binary format, regardless of the base. A well proved phenomenon (except for unary, and base infinity, to a certain extent). |
|
| |
This is not about efficiency of storage and on the four-bit level this is still binary. There are no individual storage units with twelve states. It's about how easily operations can be done on the data. As you noted, storage space is not currently an issue. |
|
| |
If it is about operations on the data, why worry about the format ( I thought it was data storage)? The machine calculation will be in binary form, why add a conversion to binary form for the calculation? Or do you presume a duodecimal calculation (including all the gaps) will be optimal. Binary operations don't have gaps. |
|
| |
Thanks, [bigsleep].
[4whom], i think it'd be quicker and easier to conceptualise, so the algorithms for manipulating it will be simpler. |
|
| |
Yes! Excellent idea - instead of RGBA or CMYK, you
want a colourspace that includes a Polarisation angle
(0-360) for linear polarisation, and additionally
something I'm yet to properly understand that will
measure "spiralised" polarisation values, like a
mantis shrimp. |
|
| |
Oh yes, my favorite animal (for this exact reason)! |
|
| |
I haven't heard of "spiralized polarization" before. Does it
mean circular or elliptical polarization, in which cases the
Poynting vector traces out a helical path (which some
people might (erroneously, I'd say) call a spiral)? Mantis
shrimp are known (to me, at least) to be able to see that
stuff. |
|
| |
To discriminate circularly polarized light with a camera,
I'd take the approach of first converting it to linear
polarization using a quarter wave plate, and then filtering
it with a linear polarizer. That wouldn't work, I know, but
it would be the first thing I'd try. |
|
| |
//(which some people might (erroneously, I'd say)
call a spiral)?// Erroneously is my middle name! Yes,
circular, helical, and "spiralized" polarisations are all
synonyms in the zen_tom personal thesaurus. I really
ought to get around to publishing that somewhere for
reduction of confusion in my general wake - it's one
reason I've come to the conclusion that I was never
really cut out for communication forms that rely on
words. |
|
| |