Re: 25 Reasons to Aviod the SD-10 (was 15 Reasons to Aviod the SD-10)

I thought you said you did translation for a living.
Do I have the wrong person?
If this is you, aren't you supposed to *understand* what you
translate?

We are NOT speaking of a tube of water, or cars, or anything but the
X3 Foveon sensor.

But, to answer your question, the thermometers would be sampling
*different* water, which could indeed be at different temps at
different depths.
With the X3, the sensors are samling the *same* light at 3 levels,
minus filtering in the silicon.
There is no attempt to differentiate any special characteristics of
the light *except* the color.

Bill Funk
Change "g" to "a"

 

Huh? What do you mean by this?

Different wavelengths of light penetrate to different depths, but all
the light reaching the 3 levels of photosensors comes from the same
position in the image plane, and that in turn comes from the same
objects in the original photographic subject. Thus, all levels of
sensors in one location, no matter how many they are, form a single
image pixel.

Everyone in this discussion has always agreed that the three
measurements taken at one photosite are spatially separated in silicon
depth. We will get nowhere as long as you insist on arguing about what
everyone already agrees is true.

Because that's the essence of the definition of pixel! The meaning is
simple: if it comes from a spatially-separated place in the 2D image
plane, then it's part of a different pixel. If it differs in colour, or
depth, or *anything* else, but comes from the same 2D location in the
image, then it's part of the same pixel.

It is counted - it just doesn't get to be named a pixel. Every molecule
in your body is doing something or other, but the individual molecules
are not each named Laurence.

That's nonsense. All the measured data is used. If you start parroting
George like that, your credibility drops as low as George.

Isn't it rather stupid to use a description of the Bayer process that
you know isn't true? Particularly when it's the same rubbish that
George writes? It doesn't do any good to say "don't get on my case"
about it - if you say stupid things, you look stupid regardless of
whether I comment or not.

One photosensor is *not* the smallest addressible element of an image.
It's merely 1/3 of the structures that capture the data that forms a
pixel of the image.

Not by any of the existing definitions. You're discarding them and
creating your own incompatible definition. That is what I do not
accept.

No they aren't. They are not spatially distinct in the 2D image plane.
By your argument, each separate colour component (red, green, and blue)
at every location should itself be called a pixel. But they are not:
red, green, and blue together form one pixel, not three.

I did not mention mosaic filter sensors at all in that paragraph. Where
do you get this from? The Bayer array is one of the few examples of a
sensor that does *not* make 3 measurements for each output pixel,
instead calculating colour using clever signal processing. Still, it's
clear that the number of spatial measurement locations equals the number
of pixels for that sensor.

Dave

 

In message <>,

You can't be serious. In the tub, you are measuring the temperature at
different depths. In the foveon sensor, what you are measuring is not
color *at* different depths; you're measuring color in a 2D grid *by*
the depth at which they are absorbed. The structure of the sensor uses
depth to measure levels of three wide color bands that are all focused
or unfocused together at the surface of the sensor.
--

 

In message <>,

What a joke - you just want to bend terminology to award the
Sigma/Foveon cameras with a semantic crown it does not deserve.

You are enamored with the Camera's output and/or theory, and want to
give it an award - more pixels than it actually has.
--

 

In message <>,

None of the definitions imply 3D positioning of elements.

Even if they did, the Foveon does not measure the color present *at*
different depths; it measures the light intensity in a 2D plane, and
only uses depth as a discriminator to divide the light into 3 wide color
bands. The depth is most definitely *not* part of the image.
--

 

In message <[email protected]>,

Huh?

There is no way to distinguish "focused" vs "non-focused" at different
depths in the Foveon sensor. Light has either entered the stack, or it
has not. All focusing occurs at and before the surface.
--

 

<grin> I'm imagining 3.4 million tiny gnomes in every SD9, each equipped
with a microscopic slide-rule & a bucket.

 

Agreed - but the key word here is *potential*. Foveon doesn't do this,
so it's rather pointless even mentioning it. In any event, you couldn't
extract any useful depth information as the difference in focal position
is *far* too small! It must be fractions of a milimeter, whereas the
depth information we'd be interested in is of the order of metres. And,
unless we had a coherent light source, you couldn't use phase
differences either. So, I stand by my comment - the Foveon has no 3D
application whatsoever.

 

No they don't, they measure sort of red, sort of green and sort blue,
however the wavelength's sensed at each sensor are hardly perfect, and
sometimes overlap, mainly due to inconsistencies in the doping process
and in the silicon manufacturing process. The consequence is that the
Foveon data has to be interpolated at every site. One only needs to
look at the unprocessed raw data from a Sigma camera, it has muted
colours mostly in the brown/orange and gray green spectra.

Your BS about quantum mechanical accuracy is simply that, BS

GK

 

Hell, every single pixel is interpolated, have you ever seen the
unprocessed output from a Foveon sensor, it doesn't resemble real
world colours at all, they are muted and tend towards gray-green,
brown and orange.

GK

 

it's still a million, no matter how you try to redefine it.

GK

 

Kibo informs me that (Laurence Matson) stated that:

"Smallest addressable point on a digital image", to be precise. I notice
that you keep on misquoting this. I take it that English wasn't one of
your best subjects at school?

That's because you seem to be exceptionally stupid.

FYI, the 3 dimensional equivalent to the term 'pixel' is the term
'voxel', something I've pointed out to you several times already. A
pixel is, by definition, 2 dimensional - a fact that's also been pointed
out to you quite a few times.

"Smallest *addressable point*", Laurence, which has also been pointed
out to you any number of times. The R, G & B layers in the Foveon sensor
are not "addressable points".

 

Kibo informs me that (Laurence Matson) stated that:

Laurence, these facts have been explained to you patiently & clearly by
myself & numerous other people in this group. You have clearly gone way
past ignorance, & have descended into pure trolling. Accordingly, I'm
not going to waste any more of my time explaining the basics of digital
imaging to you, when there are beginners asking questions here who're
actually interested in learning something, rather than shilling for
their patron.

 

SNIP

I was just humoring you, unless you mean, that wasn't what you hoped to
provoke with your condescending tone? In that case I'm sorry, but I do
wonder what else you are trying to achieve with it?

Bart

P.S. It won't hurt to snip remarks or questions you choose not to react to,
it'll save bandwidth.

 

SNIP

How hard is it to accept that in the context we're dealing with here, an
image/picture is the result of a rectilinear projection in a 2-dimensional
plane?

Bart

 

Unlikely, because e.g. X-ray tomography was a common technology even before
digital imaging existed. It was made a lot easier to perform with the the
advent of Computed Tomography, but that all deals with voxels (three
dimensional pixels).

Given the angle of projection, the refractive index of Silicon, and the
close proximity of the spectral sampling bands, it is irrelevant for the
subject at hand, pixels in a 2-dimensional plane.
The same applies to projections on photosensitive film or paper.

That's not correct. Locations in the projection direction are not spatially
discrete in the image plane.

SNIP

They're called (photo)sensors, and have been since their invention.

There's no need to redefine a nonexisting thing. Pictures are 2-dimensional,
picture elements are therefore also 2-dimensional, how hard to accept is
that?

Bart

 

Laurence Matson wrote:

"We" absolutely do not need to redefine pixels. But thanks for offering
a choice.

 

You seem to be confusing "data" for "image".
The data that come from the X3's sensors are not an image; they are
interpreted, or processed (dare I say 'interpolated'?) into an image.
As such, since the data from, say, the bottom sensor in any given
sensel location on an X3 sensor is not a part of an image (because the
image has not, at this point, been processed), it can't be a pixel.

You like analogies; that data is analogous to an egg in a cake mix
before it's baked into a cake. The egg isn't part of the cake, because
the ingredients haven't been processed into a cake yet.
Once the mix has been baked into a cake, the egg ceases to exist as an
egg, and is now part of the cake.
In like manner, that data from the 'red' sensor isn't a pixel, it's
part of the data that goes into making a pixel. Once the data from the
three sensors is processed into a pixel, that 'red' data ceases to
exist on its own.

Bill Funk
Change "g" to "a"

 

You are confusing the image with the acquisition of the data that is
processed to form the image.
The individual sensors are not an image; the data they collect is not
an image. It is *processed* into an image.

No, we just need to understand the processes involved, and to
understand the current definitions of words.
You say you understand that he definitions of 'pixel' refers to
*images*, yet still continue in your belief that the sensors comprise
an image; they don't.

Bill Funk
Change "g" to "a"

 

You are probably correct - it cannot in practice be used.

But let us stretch our minds a bit - just for the sake of
an argument. Let us assume that the three layers are at
a more significant distance. Then, different things
in the scene will be in focus at different layers. If
they are in focus or not depends on their distance. So,
the contrast can be use to meassure distance. In fact, this
is (more or less) how many auto focus mechanisms work.

In this case the distance difference is too small to be
usefull - probably. But, there is a potential.


/Roland