A white light source will produce different colors in a thin film
depending on the angle of the light source.
So, by simply sandwiching a flat retina between a thin transparent
layer and a reflector, one could get a very simple eye that's
incapable of imaging but which can use color to detect the angular
position of a light source (or the source of a shadow).
It would, however, be subject to confusion when looking at brightly
colored objects- especially other iridescent objects, which might
appear to jump around the visual field as their apparent color
changes.

So, I wonder if there is any function or environment in which that
would not be a fatal flaw. Is there some reason that an alien or
genetically engineered creature might rely on an iridescence eye?

On Feb 2, 8:50 pm, Logan Kearsley <chronosur...@gmail.com> wrote:
> A white light source will produce different colors in a thin film
> depending on the angle of the light source.
> So, by simply sandwiching a flat retina between a thin transparent
> layer and a reflector, one could get a very simple eye that's
> incapable of imaging but which can use color to detect the angular
> position of a light source (or the source of a shadow).
> It would, however, be subject to confusion when looking at brightly
> colored objects- especially other iridescent objects, which might
> appear to jump around the visual field as their apparent color
> changes.

I don't think this idea can work unless there's only one light
source, and it's more or less a point light source. Also,
in order to get any information at all the color sensors need
to be filtered to only receive narrow bands of frequencies.
That means severely limited sensitivity. The way our color
vision works, each light receptor accepts a relatively broad
band of frequencies. This makes our vision more sensitive
to differences in brightness rather than differences in color,
but it also makes our eyesight more sensitive.

> So, I wonder if there is any function or environment in which that
> would not be a fatal flaw. Is there some reason that an alien or
> genetically engineered creature might rely on an iridescence eye?

Even assuming the iridescence eye works just fine, I don't
see how useful the information is. Knowing a light source
is along the surface of some cone doesn't seem as useful
as, say, knowing whether the light sources is vaguely to the
left or vaguely to the right. The latter information may not
be precise, but it's enough to crudely hunt down or avoid
the light source. With the information your sensor provides,
you can't even do that.

Isaac Kuo

On Feb 4, 9:49 am, IsaacKuo <mech...@yahoo.com> wrote:
> On Feb 2, 8:50 pm, Logan Kearsley <chronosur...@gmail.com> wrote:
>
> > A white light source will produce different colors in a thin film
> > depending on the angle of the light source.
> > So, by simply sandwiching a flat retina between a thin transparent
> > layer and a reflector, one could get a very simple eye that's
> > incapable of imaging but which can use color to detect the angular
> > position of a light source (or the source of a shadow).
> > It would, however, be subject to confusion when looking at brightly
> > colored objects- especially other iridescent objects, which might
> > appear to jump around the visual field as their apparent color
> > changes.
>
> I don't think this idea can work unless there's only one light
> source, and it's more or less a point light source. Also,
> in order to get any information at all the color sensors need
> to be filtered to only receive narrow bands of frequencies.
> That means severely limited sensitivity.

It just requires an extremely high receptor density. The problem might
be mitigated by using sensors that receive multiple overlapping
frequency bands, so that color can be determined by what combinations
of sensors fire.

> > So, I wonder if there is any function or environment in which that
> > would not be a fatal flaw. Is there some reason that an alien or
> > genetically engineered creature might rely on an iridescence eye?
>
> Even assuming the iridescence eye works just fine, I don't
> see how useful the information is. Knowing a light source
> is along the surface of some cone doesn't seem as useful
> as, say, knowing whether the light sources is vaguely to the
> left or vaguely to the right. The latter information may not
> be precise, but it's enough to crudely hunt down or avoid
> the light source. With the information your sensor provides,
> you can't even do that.

You *can* do that. Any single point of color will give you a cone, but
there will be an arc of constant color concave towards the light
source with the colors varying differently on each side depending on
whether the angle is increasing or decreasing. So, by combining
information from lots of pixels across the whole eye, you can get
directional information with resolution limited by the number of
different colors you can distinguish.

Thinking on it some more, though, I'm not entirely sure how one would
deal with the superposition of multiple light sources muddying the
pattern. I shall have to ponder this, but for now it seems like the
best you could do is to make the eye fairly small, so you don't even
try to get differential color information across a large surface, and
average color information over the whole surface so that you can say,
for example, lots of red means theres a light source somewhere on
*this* cone, and lack of blue means there's an obstruction somewhere
on *that* cone.

Multiple eyes spaced apart and facing in different directions could
still give you multiple cones, though, so that you can deduce that a
light source or obstruction must lie on the intersection of several
cones. Hm. And I don't see why said 'multiple eyes' couldn't just be
different regions of a single larger eye surface.... Aha! And that
gets you to a single (or two, or maybe three- one on each side of the
animal) large eye surface that can use color information to determine
the angular position of light or shadow sources.

-l.

On Feb 4, 11:04 pm, Logan Kearsley <chronosur...@gmail.com> wrote:
> On Feb 4, 9:49 am, IsaacKuo <mech...@yahoo.com> wrote:
[...]
> > Even assuming the iridescence eye works just fine, I don't
> > see how useful the information is. Knowing a light source
> > is along the surface of some cone doesn't seem as useful
> > as, say, knowing whether the light sources is vaguely to the
> > left or vaguely to the right. The latter information may not
> > be precise, but it's enough to crudely hunt down or avoid
> > the light source. With the information your sensor provides,
> > you can't even do that.
>
> You *can* do that. Any single point of color will give you a cone, but
> there will be an arc of constant color concave towards the light
> source with the colors varying differently on each side depending on
> whether the angle is increasing or decreasing. So, by combining
> information from lots of pixels across the whole eye, you can get
> directional information with resolution limited by the number of
> different colors you can distinguish.
>
> Thinking on it some more, though, I'm not entirely sure how one would
> deal with the superposition of multiple light sources muddying the
> pattern. I shall have to ponder this, but for now it seems like the
> best you could do is to make the eye fairly small, so you don't even
> try to get differential color information across a large surface, and
> average color information over the whole surface so that you can say,
> for example, lots of red means theres a light source somewhere on
> *this* cone, and lack of blue means there's an obstruction somewhere
> on *that* cone.

Addendum- if all else fails, one can narrow things down by simply
wiggling the eye back and forth in order to get fairly large changes
in angles, and see which directions the colors vary in. Or stretching
the eye over a bump or depression rather than a flat surface, to get
the same result without having to move.

> Multiple eyes spaced apart and facing in different directions could
> still give you multiple cones, though, so that you can deduce that a
> light source or obstruction must lie on the intersection of several
> cones. Hm. And I don't see why said 'multiple eyes' couldn't just be
> different regions of a single larger eye surface.... Aha! And that
> gets you to a single (or two, or maybe three- one on each side of the
> animal) large eye surface that can use color information to determine
> the angular position of light or shadow sources.
>
> -l.

On Feb 5, 12:04 am, Logan Kearsley <chronosur...@gmail.com> wrote:

> Thinking on it some more, though, I'm not entirely sure how one would
> deal with the superposition of multiple light sources muddying the
> pattern.

I don't think it's possble. It's similar to the problem of de-
blurring a
photo which has suffered a blur operation. In the case of gaussian
blurring, the information is provably lost forever. With this more
complex "rainbow" blurring, more information might be retained,
but I'm skeptical about recovering it.

Isaac Kuo

On Feb 5, 1:06 pm, IsaacKuo <mech...@yahoo.com> wrote:
> On Feb 5, 12:04 am, Logan Kearsley <chronosur...@gmail.com> wrote:
>
> > Thinking on it some more, though, I'm not entirely sure how one would
> > deal with the superposition of multiple light sources muddying the
> > pattern.
>
> I don't think it's possble. It's similar to the problem of de-blurring a
> photo which has suffered a blur operation. In the case of gaussian
> blurring, the information is provably lost forever. With this more
> complex "rainbow" blurring, more information might be retained,
> but I'm skeptical about recovering it.

I'm pretty sure I've got it figured out. Each light source or shadow
source will produce a set of arcs of pure colors (or lack of colors
for shadow sources). When you superimpose multiple light sources and
look at it with a human eye, we perceive composite colors that jumble
up the information- but the iridescence eye doesn't have to do that.
It can have, say, a red arc intersecting a blue arc and process those
separately rather than looking at it as a purple blob.
Our retina's have built-in neural circuitry for picking up straight
lines at various angles- the iridescence eye retina just needs
circuitry for recognizing radius and direction of curvature of
monochromatic arcs.
An arc of any particular color can tell you the conical angle via
color, direction by concavity, and range by radius of curvature, for
objects at a particular angle. Adding lots of different types of color
sensors that gather arc information independently from other types
allows you to look for things over a range of angles. Then later
integrating arc information from each different sensor set allows you
to refine direction and range estimates by examining the ordering of
colors in concentric arcs and the spacing between the arcs.

The only problem I can think of is color jamming- very brightly
colored objects would eliminate some of the arcs, but I think that can
be worked around as long as you have sufficiently fine frequency
resolution. Possibly more problematic would be other iridescent
objects, which might seem to jump back and forth as their apparent
colors change.

-l.

On Feb 5, 2:33 pm, Logan Kearsley <chronosur...@gmail.com> wrote:
> On Feb 5, 1:06 pm, IsaacKuo <mech...@yahoo.com> wrote:

> > On Feb 5, 12:04 am, Logan Kearsley <chronosur...@gmail.com> wrote:

> > > Thinking on it some more, though, I'm not entirely sure how one would
> > > deal with the superposition of multiple light sources muddying the
> > > pattern.

> > I don't think it's possble.  It's similar to the problem of de-blurring a
> > photo which has suffered a blur operation.  In the case of gaussian
> > blurring, the information is provably lost forever.  With this more
> > complex "rainbow" blurring, more information might be retained,
> > but I'm skeptical about recovering it.

> I'm pretty sure I've got it figured out. Each light source or shadow
> source will produce a set of arcs of pure colors (or lack of colors
> for shadow sources). When you superimpose multiple light sources and
> look at it with a human eye, we perceive composite colors that jumble
> up the information- but the iridescence eye doesn't have to do that.
> It can have, say, a red arc intersecting a blue arc and process those
> separately rather than looking at it as a purple blob.

You simply won't get distinct arc shapes unless there's just
one or a small number of point light sources. That is not what
real world imagery looks like. In the real world, you don't have
point light sources, you have a continuous image.

Instead of distinct arc shapes, you just get a blurring effect.

> Our retina's have built-in neural circuitry for picking up straight
> lines at various angles- the iridescence eye retina just needs
> circuitry for recognizing radius and direction of curvature of
> monochromatic arcs.

Nope, not good enough. You won't get monochromatic
arcs. Instead, you'll get blurs. For example, consider
what a solid disc (like the Sun) will look like, compared
to what a hollow circle will look like. Both will look like
a blurred dot, with some tinting due to the rainbow effect.
If the diameter is very small, then you'll get arcs like you
expect. If the diameter is large, then you'll just get a
blurred dot with a slightly tinted fringe. Either way,
you can't tell the difference between the two images.

Now, that's what happens when you have a reasonable
approximation of a single light source. But in the real
world, you have a continuous image of objects all around.
Everything will just be a blur. You might be able to tell
the vaguely greener ground from the vaguely bluer sky,
and maybe tell vaguely where the Sun is. But you could
have done that without any iridescent stuff just as well.

Isaac Kuo

In article
<2c1ae792-8a5c-4d8b-80fe-f14246bc00bc@z17g2000hsg.googlegroups.com>,
IsaacKuo <mechdan@yahoo.com> wrote:

> I don't think this idea can work unless there's only one light
> source, and it's more or less a point light source.

It might be useful for plants (or other photovores) who need a source
of light bright enough to eat, and thus only care where the sun is.

--
David M. Palmer dmpalmer@email.com (formerly @clark.net, @ematic.com)