h a l f b a k e r yRight twice a day.
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,
|
|
|
Teleview Privacy Glass
Glass that randomly displaces rays within some radius but does not alter their direction | |
Tele from the Greek word meaning "at a distance". I call it
teleview
glass.
If a type of glass could be constructed so that it displaces a "ray"
vector randomly within some radius but does not alter its
direction
then some interesting uses exist.
Looking at stuff through the glass, everything
would be blurred by
a
fixed size radius, not an angular cone the way frosted glass does.
Lets say for example 10cm radius.
Objects closeby would be greatly blurred, people's faces
unrecognizable and it would be suitable for use in bathrooms,
showers or locker rooms.
Objects very far away, so far that the angular resolution of the
human eye is significantly more than 10cm at that distance, would
appear unblurred. So you could still have a view on mountains,
clouds or the ocean. To a lesser extent, trees.
It would now be possible to have showers, locker rooms, dressing
rooms etc. with a clear view at a distance and still full privacy.
I've already come up with one way to consruct such glass (very
thick,
expensive and need for precision), but the idea here is the
concept.
At some point I will render a sample picture in CG.
Frosted Glass
http://www.freeimagehosting.net/pty3j
Back ray-trace of regular frosted glass [Eon, Apr 08 2013]
Teleview Glass
http://www.freeimagehosting.net/ahm4g
Back ray-trace of teleview glass [Eon, Apr 08 2013]
View
http://www.freeimagehosting.net/bv159
View from behind teleview privacy glass [Eon, Apr 08 2013]
Rendered, with refraction
http://i923.photobu...ion_zps0561c338.jpg
The near, small figure (red) is scrambled, as is the far, big (blue) figure. [MaxwellBuchanan, Apr 09 2013]
Same scene, without refraction
http://i923.photobu...ion_zpsd4caa1dc.jpg
just to show there's no trickery. [MaxwellBuchanan, Apr 09 2013]
Same scene again, shown from the side. QED
http://i923.photobu...ene_zps8f270a24.jpg
The "teleview" glass is the white vertical line to the right, as seen edge-on. [MaxwellBuchanan, Apr 09 2013]
Refraction by a slab at an angle
http://i923.photobu...ion_zps2e3ef4b8.jpg
Red figure is near and small; blue is far and big. [MaxwellBuchanan, Apr 10 2013]
As above, but slab has refractive index of 1
http://i923.photobu...ion_zpsee29f94c.jpg
No refraction. The transparent slab is invisible. [MaxwellBuchanan, Apr 10 2013]
...and the whole scene viewed from off to one side
http://i923.photobu...nce_zps5ee0df75.jpg
The glass block is grey here, for visibility. [MaxwellBuchanan, Apr 10 2013]
...and top view for [MechE]
http://i923.photobu...iew_zps54c72c15.jpg
I've put a grey ball in place of the red figure, which was too small to see. [MaxwellBuchanan, Apr 10 2013]
The Impossible Lens
http://www.wipo.int...studies/frazier.htm
A bit of the story of Jim Frazier. An interesting doco. [AusCan531, Apr 11 2013]
Obligatory Father Ted link
http://www.google.c...m=bv.45107431,d.aGc
[spidermother, Apr 11 2013]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
| |
Hang on a second. If the glass displaces bits of the
image by 10cm in a random direction, and if this
"10cm" is measured on the glass (rather than on the
object being viewed), then a cow in the distance is
going to be blurred by a cow's-image-length, whilst a
naked person standing just on the other side of the
glass will be blurred by less than a person's-penis-
image-length. |
|
| |
I therefore contend that this will fail. |
|
| |
This idea teeters precariously upon the
word 'if'. |
|
| |
Welcome to the Halfbakery, [Eon]. |
|
| |
I was taking the "If" as a synopoeotic interjection. |
|
| |
My point was that, if the "if" were indeed
satisfied,
then the glass would still not fail to misperform as
described. |
|
| |
But indeed, welcome to the hollowed grounds of
the Halfbakery. In positing a material which not
only doesn't exist, but would also fail to do the
required job if it did exist, you have indelibly
marked yourself out for greatness here. |
|
| |
Or it may just be a gravy stain. |
|
| |
OK first, provided that the departure angle of rays
aren't affected by the glass, there is no reason
why a cow should be full length blurred. The blur
radius remains 10cm regardless of distance. |
|
| |
As for feasibility of construction, imagine a 1D blur
first. If this could be constructed then placing two
perpendicular would give the desired effect. One
can displace a ray by having it pass through some
thick sheet of glass at an angle. Refraction
happens twice, the second time restoring the
direction but there is a net displacement of the
vector. A 1D blur could be achieved by having a
sandwich of such glass strips (say vertical), very
thin, at various angles (tilts), each displacing the
ray by some amount (vertically). Looking through
the glass means looking at the side of
the vertical sandwich. Between the slices there
would have to be another medium. Total internal
reflection keeps light from crossing between
slices. Circular polarization on the two ends
ensures that only rays that bounced an even
number of
times pass. Lots of light will be blocked, but from
the remaining light the desired
effect would be achieved, at least in theory. I
didn't want to get into the details but I don't want
my idea to be perceived as wishful thinking. |
|
| |
Around here you have to get into details or your idea
_will_ be perceived as wishful thinking. This place is all
about getting into details. Other 'bakers will complain if
you don't. |
|
| |
So, it's kind of a deliberately made out of alignment Fresnel lens? |
|
| |
//The blur radius remains 10cm regardless of
distance.// |
|
| |
Yes, but it's the radius of the blur as measured on
the image as it passes through the plane of the
glass. |
|
| |
Suppose you look at a face, through the window.
Suppose also that both you and the other person
are about a metre away from the glass. A light ray
from their left ear to your eye, and a light ray
from their right ear to your eye, will travel
through the glass at points about 10cm apart.
Your glass randomly displaces the image by 10cm,
so their face will appear completely blurred. |
|
| |
Now you look at a distant mountain. Once again,
you're about a metre from the glass. The
mountain's size and distance are such that it
appears about as wide as the person's head did. |
|
| |
A light ray from the left side of the mountain, and
a light ray from the right side of the mountain,
both travelling towards your eye, will pass through
the glass about 10cm apart. Since your glass
displaces the light rays by about 10cm, the image
of the mountain will once again be fully blurred. |
|
| |
In other words, what's important is the *apparent*
size of the object, as projected onto the window.
A big object far away appears as large as a small
object close up, and therefore both will be
equally blurred. Still, apart from a fundamental
misunderstanding, it's a nice idea. |
|
| |
But what's with the pinhole camera? |
|
| |
Seriously, unless the glass can magically tell whether
a light ray is coming from a small thing close up, or a
big thing far away, this ain't gonna work. |
|
| |
If you think different, just sketch a diagram of the
rays coming from opposite edges of a big distant
thing and from a small near thing, to a point (the
eye) and show me how they behave. |
|
| |
One can simulate what would be seen by using
backwards raytracing.
It basically means only considering those rays that
ended up hitting your eye from a certain direction
and then asking where it collected light from by
starting at the eye and tracing its path (collection
area) in reverse.
Consider someone standing next to the window on
the outside and a big building in the distance.
With the collection area indicated in red, link 1
shows what happens with normal frosted glass.
From the observer's point of view, the amount of
blur is more or less angularly constant. The
absolute blur radius in world units increases with
distance.
Link 2 has teleview glass picture.
The absolute blur radius is constant, or from the
observer's point of view, angularly decreasing with
distance.
Link 3 shows what such an observer could see:
That last image was just constructed from layers
and pics I found on the internet, with some blur
effects, not REALLY ray-traced. But I hope that
helps explain it :) |
|
| |
I see the flaw in your reasoning - "parallel rays".
Stop and think. |
|
| |
I agree with [Eon] and [bigsleep]. It's fairly easy to see how this would work with a pinhole camera, but I'm pretty sure it would work equally well with a large aperture camera as well as long as it was focused on the distant object. And of course the catch is creating the material that can displace the rays of light without altering their angle at all. |
|
| |
[Eon], could you make a diagram of your proposed solution? I think I basically figured out what you mean, but my first impression was way off, and it's hard to provide criticism if I don't understand what you mean. |
|
| |
Does Iceland Spar have the basic required optical characteristics? It creates a double image of close objects (see picture in Iceland Spar link), yet doesn't appear to cause significant distortion of distant objects (see photo of Iceland Spar in Sunstone link). Of course making a bathroom window from these might not be practical since by the time you layered enough to get random displacement rather than double images, I suspect your clarity would be completely gone. I think this shows that such a window is theoretically possible. |
|
| |
Here's another implementation concept: Make some mini-blinds with flat slats. Each slat is a double sided half-silvered mirror. Set the blinds at 45 degrees. A perpendicular ray of light will have a 50% chance of going straight through. The other 50% will turn 90 degrees and hit the next slat. 25% will reflect off this and be offset by one slat spacing. 25% will go through this slat and hit the next with 12.5% reflecting at 2 slat spacings, etc. If viewed from 45 degrees, one way it will be transparent. At the other 45 degree angle only light reflected from inside will be visible. If there is glass in the space between the slats, the refraction will result in reduced undesired reflection. Transparency would of course be eliminated simply by having two layers with opposite angles on the slats, so you'd have at least 4 layers with different orientations to provide full privacy and 2D blurring. |
|
| |
So did you actualy understand what you just quoted any better than I did? Does that mean this it could theoretically be used for this purpose or not? |
|
| |
//do I need to do a full-on stochastic simulation// |
|
| |
No, just sketch me a picture showing: |
|
| |
(a) the observer
(b) the window, wherein light rays are displaced in
a random direction by 10cm
(c) A small arrow close to the window
(d) A big arrow far away from the window,
subtending the same angle as (c)
and
(e) The light rays reaching the observer's eye from
the head and tail of each arrow. |
|
| |
Yes, that's the sort of diagram of which I was
thinking. And it shows that the amount of blur is
the same for the near and the far object. |
|
| |
So you'll see a nearby face as an oval blob. You'll
see the distant mountain as a mountain-shaped
blob. No? |
|
| |
I'm not sure I see what you mean by your
"effective X pixel resolution" labels, though. If
you're saying that the visible resolution of a near
object is 5 pixels whilst that of a far object is 20
pixels, then you're saying that the nearby face is
*less* blurred than the distant mountain, which
can't be right either. |
|
| |
Also, on the diagram, your rays (and viewing cone)
converge to a point. What happens if you make
the situation more realistic, i.e. the retina is a
small screen placed just behind the point at which
the rays (on your diagram) converge? |
|
| |
OK, crayons and a large envelope back suggest
that your model only applies for a "point sensor"
(the point at which your light rays converge)
which has zero spatial resolution. |
|
| |
If, in place of your point-like "observer" you
instead place a lens (the eye's lens) of finite and
fairly small diameter, and if you then place a
curved screen (the retina) behind that lens, then
things fall apart. If you want to consider a ray-
tracing with the rays coming out of the eye and
hitting the object (which, agreed, gives the same
result as the actual situation of light rays going
from object to eye), then you have to consider
rays coming out of the eye over a range of angles
from each point on the retina. |
|
| |
It's more complex than I had first thought, but I
am still certain that it can't work, because the
virtual image of any large, distant object will the
the same as the virtual image of a small, near
object when taken at a plane close to (and just on
the far side of) the glass. |
|
| |
OK, I modelled it in Cinema4D. Here's what I did: |
|
| |
(1) create a hexagonal column (hexagonal
prism)
(2) Sliced the two ends of it at the same angle, so
the two end-faces are parallel to eachother, but
not at right-angles to the long axis of the
column.
(3) Gave the column a refractive index of 1.5 (for
glass). This column, viewed end on, will displace
an image as described in the idea.
(4) duplicated this column 31 times, displacing it
each time to make a row of columns just touching
on their long flat sides.
(5) Rotated each successive column by 60°. Thus,
each successive column will displace an image
upward, up-and-left, left, down-and-left etc.
(6) Duplicated this entire row of columns 31
times, displacing each row
|
|
| |
OK so far? This gives us a "bundle" of hexagonal
columns. They look like a honeycomb when
viewed end-on. However, each of the columns
will displace an image in a different direction.
This entire collection of columns is our sheet of
"teleview glass". |
|
| |
Are you comfortable that, so far, I'm complying
with the idea? |
|
| |
(7) I put a viewpoint on one side of the "glass".
Everything is now seen from an "observer" at that
point.
(8) I added a human figure, placed on the far side
of the glass and fairly close.
(9) I made a duplicate of the figure, and scaled it
up by 10-fold in each dimension.
(10) I moved the enlarged figure further back from
the glass (10-fold further from the viewer) so that
it appears the
same size as the first figure. I also moved it to
one side, so you can see the near, normal-sized
figure and the distant, giant figure side by
side.
(11) I then rendered the scene.
(12) I turned off refraction (ie, I set the
refractive index of the glass to 1.0) and re-
rendered the scene, just to show that there's no
trickery. |
|
| |
(13) I also rendered the scene from a viewpoint off
to one side, so you can see that the figures really
are different sizes and at different distances from
the "teleview" glass (which appears as a white
vertical line toward the right of the image in the
third link). |
|
| |
I'll provide a link to the images. In each case, the
RED figure is the smaller, nearer one; the BLUE
figure is the further, larger one. |
|
| |
To my eye, they are both equally scrambled. Tell
me if you think differently. |
|
| |
//The distortion is just an artifact of the column
bundle and not just ray translation.// |
|
| |
Unlikely. C4D is generally pretty good. The
columns are parallel. Their opposite faces are also
parallel (I angled the ends by intersecting with a
cube, i.e. both ends are 'cut' parallel to
eachother). The image without refraction (but
with everything else) is undistorted. But perhaps
by 'column bundle' you are thinking of something
else? |
|
| |
Short of getting this thing fabricated, I don't think
I can go any further - I'm satisfied that it won't
work, but it was an interesting problem. |
|
| |
(For the record, I also went on to try this with a
two-fold finer pattern of prismatic glass; it gives
essentially the same result.) |
|
| |
If you think differently, please feel free to go
ahead with your stochastic simulation, or a
rendering in whatever modelling software you
prefer. Probably a "pinhole lens" is fine (and will
be simpler) - just don't forget to trace through
onto the imaginary retina. |
|
| |
After point 6 I'm not happy as teleview glass can
not be constructed this way. Many of the rays
inside the hexagonal prism will undergo total
internal reflection which does way more than
displacement - those rays will end up leaving the
prism with a different angle than they entered
with. If you make the prism tubes short enough so
that this doesn't happen a lot then the
displacement will again be too short. It really isn't
easy to construct such glass, I don't think it would
be easy to simulate in an off-the-shelf ray-tracer (I
wrote a basic one in C++, if I get time over the
weekend I will generate an image). |
|
| |
Granted, my first two links show what happens for
a single "pixel" sensor and yes, I assume that that
pixel is infinitesimal. If I assumed the pixel had
some size then yes, it would be a very thin cone
that leaves the eye/camera and the region after
the glass would very, very gradually open up. But
the angle of this would be very, very tiny, like
0.01 degrees (as per angular resolution of sensor). |
|
| |
There is a spread of angles as you consider a
collection of such pixels to ultimately make up an
image, but this just adds up to the field of view.
The blur is about what gets collected PER pixel
and for this it truly consists of a collection area of
nearly parallel rays. I say nearly to be pedantic,
but really it is about as good as parallel. |
|
| |
not_morrison_rm gave me a good idea on how to
construct this. Imagine birefringent glass with two
sawtooth interfaces (at different frequencies to
counter interference patterns), the short step
interfaces are blocked with black paint. The
sawtooth patterns are relatively inverted to
ensure the same orientation of the flat regions on
opposite sides. Ignoring some percentage of
blocked light, such glass would split the image
into two displaced ones. To construct teleview
glass, have 7 or so such layers, each at random
rotation and between the layers have quarter
wave plates to convert the then linear polarized
light back to circular (or basically just not linearly
polarized) so that more levels of split-up can
occur. With 7 layers there will be 128 displaced
images summed together, probably good enough
for privacy from eyes (although it is conceivable
that someone could write a computer algorithm
that could unscramble a digital picture taken).
This whole thing is still about 70cm thick, so not
yet elegant and compact. |
|
| |
//. Many of the rays inside the hexagonal prism
will undergo total internal
reflection which does way more than
displacement - those rays will end
up leaving the prism with a different angle than
they entered with. If you
make the prism tubes short enough so that this
doesn't happen a lot then
the displacement will again be too short.// |
|
| |
In the rendering, there's no internal reflection
because there's no
interface between the prisms. They are
constructional elements, and the
material (ie, transparent, r.i.=1.5) is applied to
the whole. |
|
| |
Truly, I like the idea because it is not as simple a
problem as it first
seems, but my instinct and the simulation say no. |
|
| |
Perhaps [bigsleep] will do a rendering and find
different results. It would
be interesting if he can render a scene similar to
mine. |
|
| |
[Eon] you might be interested by fibreoptic
bundles, where the fibres
needn't be parallel (for example, some flexible
light-pipes use such
bundles, where the position of one end of a single
fibre doesn't
correspond closely to the position of the other
end). These non-parallel
bundles "scramble" an image which is presented
close to one end of the
bundle. In theory you could take many such
bundles (all fairly short) and
bundle them side by side to make "glass" that
displaced incoming rays by a
short random distance. However, it wouldn't work
for things far away,
because of course the light rays from a distant
object fan out and will
strike everywhere on the bundle face (ie, you
have to think about more
than one ray from each point on the source). |
|
| |
// Of course. But with distortion, why don't you
see large areas of red and blue ? The hex columns
seem to have lost lots of the image rather than
just moving it around.// |
|
| |
I think there's roughly as much red and blue in the
distorted image as in the undistorted. |
|
| |
//Try much smaller columns e.g. half as thick
end-on as their hex widt// They're not that far off
as it stands - they're about as wide as they are
tall. There are unlimitless possibilities (make
each one less refractive or more refractive; make
the face-angles steeper, put hair on the human
figures) but, hey. Maybe there's a model that'll
work - render me an example and I'll probably be
convinced. |
|
| |
One way to achieve the desired effect with
available resources would be to replace the
window with a camera and screen. Just give the
camera a shallow depth of field set far away. On
the other hand, if the objective is to be able to
look at the mountains whilst standing there
bollock naked, the depth of field would be
completely non-unirrelevant. |
|
| |
MaxwellBuchanan, if there is no internal reflection then
that is just as bad, because the rays will end up crossing
into other prisms and have a high probability of them
leaving from a different prism than they entered and finally
refracting at an interface which is not parallel to the first
one. Direction will not be maintained. |
|
| |
// [link]s. These are ray-traced images// Well,
OK, I'm 2/3rds of the way to being convinced. I'll be
at least 3/3rds convinced if I write my own code and
get the same answer, and probably 4/3rds convinced
if anyone can actually build this and make it work. |
|
| |
[MaxwellBuchanan], can Cinema4D do half-silvered double sided mirrors? If so, it seems like it would be pretty easy to test my miniblind implementation. (last paragraph of my Apr 08 anno) |
|
| |
.........X..
.....x......
.............
/////////
.............
......o.....
|
|
| |
x is your red person.
X is your blue person (not to scale).
o is the camera.
///// is an array of half silvered mirrors that should be at 45 degrees.
|
|
| |
This only shows one layer, but that should be good enough to validate the concept. You could actually do a very simple version with just two large slats. That should just give you a double image. If the concept is correct, then the double images of the close red person will appear to be more missaligned than the double image of the far blue person. |
|
| |
I just realized I know how to test this concept in the real world. I have noticed that at night with the curtains open, our double pane windows create a double reflection. I should be able to compare the offset between the double images of near and far objects. |
|
| |
I am not convinced, and think that there is still an
inherent assumption problem in this. Light coming
from scenery to the human eye (or a camera) is
not parallel. Viewed objects are not measured in
terms of physical dimensions, but in degrees (or
minutes) of arc. Thus, a single sensor can't tell if
the object is small and near or large and far. It
happens that, they way binocular vision works,
our eyes can interpret the data to have a better
idea, but that doesn't change the optical sense. |
|
| |
And no matter how far the object on the far side
of the glass is from you, a 10cm displacement at
the glass subtends the same angle with respect to
the sensor (your eye). So it doesn't matter how
far away the object is. |
|
| |
As far as your program, [Big], the problem is when
you set the distance zones, you didn't scale the
(flat) image the same way the eye would see it.
Remember, an object 10x as far away will appear
10x smaller. Obviously the objects themselves are
scaled because they were in the original
photograph, but the image isn't. Shrink the
portions of the image
properly and run your test again, and see what
happens. Or better yest, repeat with a simple
checked pattern or similar, with the identical
piece put at the three respective planes. |
|
| |
The 45 degree half silvered mirrors will only maintain
direction for rays that bounce an even number of times
(remember to consider non 90 degree rays) so won't work
as is. |
|
| |
But we can block everything that bounces an odd number
of times with same handedness circular polarizers on either
side. Excellent! Thus far this is the easiest construction I
have seen. |
|
| |
MechE, I think the source of disagreement boils down to
this: |
|
| |
Imagine a single, 30cm thick, 45 degree sheet of glass: |
|
| |
The sheet only extends halfway up your field of view and
you are looking at a mountain behind it, lower part through
sheet, upper part over sheet. The question is: Is there an
apparent split in the image of the mountain, a discontinuity
from relative displacement, when comparing what is seen
through the sheet and over it? |
|
| |
I'm guessing you say yes. |
|
| |
I say no, not for a far off object like a mountain, but yes
for close things behind the sheet. This can be easily
raytraced. |
|
| |
And I'm saying optics don't care about how far off the
object is, just how much angle of view it covers,
so... |
|
| |
Yes, there is a discontinuity. |
|
| |
The main objection to my raytracing through an
array of prisms was that some rays will be
refracted between prisms, therefore being shifted
in angle as well as being translated. |
|
| |
I could (had I the energy and patience) redo it,
but give each hexagonal prism black side-walls.
Then no light will pass between prisms, and only
translation will occur. This would be no good for a
real window (you could only look through it at one
angle), but would do for p.o.c. |
|
| |
Anyway, I tried it with the same setup as before,
but now with each hexagonal prism bounded (ie,
divided from the others) by zero-reflectance
black. |
|
| |
The results are inconclusive. This time, the
*nearer* figure is noticeably clearer than the
further one; they both have individual "cells"
translated relative to their unrefracted positions. |
|
| |
However, I can't hand-on-heart swear that C4D is
handling the refraction perfectly. It would have
to refract each ray in an identical but opposite
way on entering and leaving each hex cell. My
guess is that, in making its angular calculations, it
will round off at some finite limit, in which case
the ray won't be restored to its original direction
after passing through the prism. |
|
| |
Under the circumstances, I think the best answer
will be from [bigsleep]'s software, since the
direction of the ray can be guaranteed to be the
same after refracting through the prism. |
|
| |
However, I'm still not certain that tracing rays
from (or to - either way) a single point-like camera
isn't screwing things up; and I can't be certain that
[bigsleep]'s software takes all relevant factors into
account. There are weird optics happening here,
and I am not convinced that the usual "point like
observer" is valid. But I am open to being
persuasioned and convincified. |
|
| |
One thing that has emerged from the simulations:
if this type of window is to be made, it would
have to guarantee zero angular deflection after
the displacement of the rays. Any residual angular
deflection is going to make the distant image
much worse than the near image - the exact
opposite of what is wanted here. |
|
| |
MaxwellBuchanan, would you be able to do that simple
raytrace I described with a single thick 1.5 refractive index
glass sheet at 45 degrees (near left to far right) that only
extends halfway up the two colored bodies? |
|
| |
I would like to see the displacement discontinuities
compared for the two bodies. |
|
| |
//a single thick 1.5 refractive index glass sheet at 45
degrees// Well, my old headmaster was right. I'll be
buggered. See links - collapse of stout party. |
|
| |
The damn thing is counterintuitive, is what it is
counter to. |
|
| |
I still reckon you'll struggle to implementize it,
though. |
|
| |
Okay, I'm having trouble explaining this, but it
cannot work the way it is being described. Let's
think about what happens if you offset the light
rays uniformly, say with a prism. The light rays
that come from an object that takes up five
degrees of the field will offset exactly the same
amount, regardless of whether that object is near
field or far. If you randomize that offset, each
pixel (for want of a better word ) will offset that
same distance in a random direction. Thus each
object will distort the same amount regardless of
how far away it is, because it subtends the same
portion of the view. |
|
| |
Far field objects don't have more pixels just
because they are far away. |
|
| |
[MB] Can you show a top view? I think I may see the
problem, but I'm not certain from the views available. |
|
| |
//Can you show a top view?// |
|
| |
Done, see link. I replaced the red figure (which was
too small to see easily from the top) with a grey
ball. The viewer (not shown) is at the bottom of the
image, in the middle, looking upwards as drawn
here. |
|
| |
Yeah, I was thinking of that and wondering if it uses
the same principle, or principal. Or both. |
|
| |
[MB]- So, the problem with that representation is
that the lens is closer to the viewer in one case
than the other. That will produce a difference.
Repeat the experiment with two identical lenses,
one for each figure, and see what happens. |
|
| |
[Big] Done, and it still doesn't work out like this.
An object that is twice as far away but subtends
the same degree of the field of view will be the
same apparent size at the optics, and thus
experience the same apparent distortion. |
|
| |
The sheet in [MB]s ray trace is a lens, so I referred to
it as such. |
|
| |
You'll notice in regard to your work, I used the more
generic optics, which is correct. |
|
| |
I think I see what you're trying to say, and it involves
treating the offset as a binocular baseline. I'm not
convinced that it works, and I suspect it's impossible
to build, but other than that you may have a point. |
|
| |
This is a fascinating discussion which I don't feel qualified to participate in, but would advise those others interested to watch the NatGeo documentary called "The Impossible Lens" about an Aussie named Jim Frazier. He came up with a lens with infinite depth of field after being told it was an impossibility. |
|
| |
//Far field objects don't have more pixels just because they are far away.// |
|
| |
I think the basic explanation of this is that they do - if you're not comparing like-for-like, but instead what you actually see nearby vs at a distance. |
|
| |
That is, if you imagine that everything is coloured in voxels, then a distant mountain will have many voxels per degree subtended at the observer, while a nearby object like a flower or human will have only a few.
Olay, so now if we consider that there is detail at all scales; that is if we look at a view we see nearby flowers in some detail, but on a distant mountainside we're seeing whole woods, the snow-line and so on - with the data from many trees, flowers rocks etc averaged together in each phtoreceptor. |
|
| |
//So, the problem with that representation is that
the lens is closer to the viewer in one case than
the other// |
|
| |
Ah, OK, you mean because the viewer is looking
through the "nearer" part of the glass sheet at one
figure, and through the "further" part for another? |
|
| |
If that's what you mean then, no, that's not the
problem. I just replaced the two figures with
vertical rods (one red, one blue), and lined them
up so that the nearer (red) rod obscures the
further (blue) rod, as seen with no refraction.
Then I turn on refraction and re-render; the
nearer (red) rod is now 'broken' (ie, the part of it
behind the glass is displaced sideways), whilst the
further (blue) rod is not. |
|
| |
I hate to say it, but I think the basic idea here is
correct after all. In one sense it's counterintuitive
(how does the glass "know" a light ray is coming
from a distant as opposed to near object). In
another sense it makes sense: the real and
displaced rays (traced backwards from the eye)
form
parallel lines which converge (ie, zero apparent
displacement) at infinity. |
|
| |
At this point I have to agree on the concept
working. |
|
| |
Basically, by having an offset between the input
and output, you get a non-zero optical baseline,
gaining the benefits of (pseudo) binocular vision
at each point. |
|
| |
Now, however, I have to start quibbling about
whether a material can exist that can do this. In
order for it to work, a photon entering at a given
point has to exit at a point between x and -x. If
this exit is at all deterministic (as it would be from
the fiberoptic version mentioned in annos), you
would not get the advantage of a non-point
optical baseline, as each point input would be a
point output, and thus the distortion at a given
point wouldn't care about the distance. |
|
| |
I think the problem with any fibre approach is that
direction won't be preserved; a ray might enter a
fibre at some angle (up to the acceptance angle of
the fibre), and will be displaced by however much
the fibre shifts on its path through the "window",
but the ray will exit the fibre at a different angle. |
|
| |
It perhaps could be done by having an array of
cameras, lensed in such a way as to have a narrow
and parallel field of view, and each linked to a
small display, but with the display of one camera
not falling behind that camera itself. |
|
| |
One-way glass would work better. |
|
| |
It would, if such a thing existed. |
|
| |
A "one way mirror" is just a piece of semi-silvered
glass. When you stand on the more brightly-lit side
of it, your reflection dominates the image you see.
To reverse a "one way mirror", just reverse the
intensities of the lighting on either side. |
|
| |
//When you stand on the more brightly-lit side of it,
your reflection dominates the image you see. // |
|
| |
Yes, as the inside of a house is darker that the
outside during the day. |
|
| |
At night you would cover the window with "shades"
or "blinds" or "curtains," all baked concepts; you
wouldn't need to look outside at night, anyway,
because it's generally too dark to see. |
|
| |
// as the inside of a house is darker //
except if it's a day with dark clouds and you have the bright lights around the bathroom mirror turned on. And even during a normal day, you'll have some rather annoying reflections reducing the quality of your view. |
|
| |
// you wouldn't need to look outside at night, anyway, because it's generally too dark to see. // That all depends. A person living in high-rise in the city would probably have a very nice city-scape to view at night. Other people might occasionally have a nice view of the moon. |
|
| |
//the window with "shades" or "blinds" or "curtains, |
|
| |
I'm just using big sheets of cardboard, with a horizontal gap at the top. That way (in theory) I can stargaze while lying in bed. But, too much cloud in the UK. |
|
| |
//except if it's a day with dark clouds and you
have the bright lights around the bathroom mirror
turned on.// |
|
| |
//A person living in high-rise in the city would
probably have a very nice city-scape to view at
night// |
|
| |
A person who can afford a high-rise apartment
with a nice city view can afford the outrageous
cost of the teleview glass. I think you found the
target market: Donald Trump. |
|
| |
So [whlanteigne] are you're still contenting that a half-silvered mirror with manual shades is "better", or are you just saying that the idea is highly unlikely to be implemented any time in the forseable future at a price that makes it a better value than standard half-silvered glass and blinds? |
|
| |
I agree with the later. I'd probably only pay a 20% premium to get this instead standard glass, which makes this a perfectly halfbaked idea. |
|
| |
[MB] Even with cameras (and why do you bother
showing anything outwards with the cameras),
you still have the problem that each camera is
monocular at the point of incidence. If you had a
series of cameras with the images overlaid, you
could get the sort of near field blurring this
describes, but, again with cameras why bother,
just display only on the direction you want to. |
|
| |
Again, I do not believe this constructable with
simple optical elements. |
|
| |
// I do not believe this constructable with simple
optical elements// Oh, I dunno. Test-drive a pair of
middle-aged eyes sometime. Stuff that's too far
away to be relevant: clear as kodak. Stuff that's
close enough to interact with: fuzzier than a
Lithuanian's armpit. |
|
| |
That's nice, but neither of those implement the idea
as described. |
|
| |
An array of spinning micro mirrors or prisms such
that each adjacent unit is scanning a different part
of the viewable area would approach the idea, but
still not the sort of uniform blurring that the idea
suggests. That effect is what I think is not possible
to produce. |
|
| |
//So [whlanteigne] are you're still contenting that
a half-silvered mirror with manual shades is
"better"// |
|
| |
Not "better," but it works, it's inexpensive, and I
can go out and buy it today. In fact, there are
films I can buy to apply to the glass to do the same
thing, so I don't have to replace the windows. I
just have to clean them well and apply the film
carefully. |
|
| |
My bathroom window has a frosty-ish film on it
that effectively blurs the view for anyone looking
in at night; there is a strategically placed,
discreet, 1/4" section cut out so I can peek out if I
need to. |
|
| |