Image processing - photo realism
Lew
like to ask for opinions based on the comment made on 60 Minutes last
Sunday, June 1. The segment was a rebroadcast of "Hailing the Hubble."
NASA imaging specialist Zoltan Levay "uses scientific data to apply
color to Hubble's black and white images." On the face of it, I've got
no problem with that. Then Levay says, "We do adjust the color a
little bit, partly just so it looks better, and partly so it also
imparts the information that we'd like to get across. It's a
representation of reality, just as any photograph is not a literal
reality, but a representation of reality."
Perhaps Levay was not accurately describing the intent of the image
processing, but he as put it, "just so it looks better," I feel it
suggests far too much subjectivity. Put another way, does National
Geographic "enhance" astrophotos (and other photos?) just to be
attractive to readers?
I am an amateur visual astronomer, no photography involved, but in my
work, I use Photoshop regularly, and I know the capabilities of such
software. I am concerned that image processing is going way beyond the
goal of scientific valid presentations and seriously distorting
reality. Is the public being given the wrong impression of what
celestial objects would look like to human spacefarers of the future?
Am I making much ado about nothing, or is this a valid ethical
concern? Does anyone care?
Bill Foley
this is NOT what the objects look like in a scope or in a hypothetical
spaceship from near up. Almost from its inception, photography has required
some corrections, such as filtering, dodging, tilting lenses to correct
perspective, and so forth - photos do need this sort of control to look natural
or they would be too contrasty, have messed-up color, would over or underdo
perspective effects, etc. Don't get your vowels in an uproar over this.
Clear, Dark, Steady Skies!
(And considerate neighbors!!!)
Hig Hurtenflurst
equalization" with their telescopes. The concept is that you use a visual
magnification that corresponds to the light magnification of the objective.
(i.e. if your objective magnifies light by 50X, then you should use an
eyepeice that gives you 50X visiual magnification).
While I by no means limit myself to "visual equalization" while
observing, I do like the idea behind it. It does give you an idea of how
things "really" look.
Tom
Alexander Avtanski
I don't think that this is a valid concern. The image processing
don't introduce "false" information in the image. Also, I don't think
that the images should represent the "natural" look of the object, i.e.
how the object would look like if the observer was looking through the
scope with his own eye. Let's say you have taken an image in
ultraviolet - what should the picture look like? An empty black
square, because ultraviolet is not visible? So, nothing wrong with
substituting UV with "blue" or "green" or whatever on the image.
The same is with enhancing colors, combining multiple exposures, etc.
- Alex
Brian Tung
> I don't think that this is a valid concern. The image processing
> don't introduce "false" information in the image. Also, I don't think
> that the images should represent the "natural" look of the object, i.e.
> how the object would look like if the observer was looking through the
> scope with his own eye. Let's say you have taken an image in
> ultraviolet - what should the picture look like? An empty black
> square, because ultraviolet is not visible? So, nothing wrong with
> substituting UV with "blue" or "green" or whatever on the image.
> The same is with enhancing colors, combining multiple exposures, etc.
Although I agree with the substance of what you've written, I think
that some kinds of image processing certainly can introduce false
information into the image, in the sense that certain artifacts appear
to the human eye. Yes, no data has been manufactured as such, but
in the effort to sharpen an image, sometimes the processing produces
a double fringe to a contour, for example, which is not present in the
original object.
I think that most changes in the color distribution won't hurt too
much, but I don't have a lot of experience with that.
Brian Tung <br...@isi.edu>
The Astronomy Corner at http://astro.isi.edu/
Unofficial C5+ Home Page at http://astro.isi.edu/c5plus/
The PleiadAtlas Home Page at http://astro.isi.edu/pleiadatlas/
My Own Personal FAQ (SAA) at http://astro.isi.edu/reference/faq.txt
Chris L Peterson
While there are processing functions (like deconvolution) that can add
structural artifacts to an image, simply adjusting color weights to achieve some
appearance does not. If National Geographic modifies terrestrial photographs in
such a way as to deviate from reality, this is apparent to anyone comparing the
image to the actual scene. But it is much more difficult to define "reality"
when you are talking about astroimages. The objects being imaged would not show
color (or very little color) to the human eye no matter how close they were
viewed from, and the method of constructing "color" images results in an
entirely different color space than the eye/brain uses, and often includes
wavelengths where the eye is insensitive. The Hubble images (and the data from
many other telescopes, too) contain scientifically valuable information in the
individual frames collected through each filter, and aesthetically pleasing
information when those frames are combined and some sort of color space is
mapped. I see no ethical problems here at all.
_________________________________________________
Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com
Jon Isaacs
>many other telescopes, too) contain scientifically valuable information in
>the
>individual frames collected through each filter, and aesthetically pleasing
>information when those frames are combined and some sort of color space is
>mapped. I see no ethical problems here at all.
>
>________________________________
So where does the science stop and the art begin?
Maybe each photo ought to have a disclaimer pointing out that the color choices
are arbitrary and have no real relationship to the frequencies of the
electro-magnetic spectrum as the human eye would see them?
Back when Rod and the rest of us were attending ZZtop concerts and the like,
there were ways to get interesting and aesthetic star cloud images without
going to the trouble of firing up the Hubble....<g>
Just wondering......
jon
Chris L Peterson
>So where does the science stop and the art begin?
Most of the science uses the individually filtered frames, not the combined and
color mapped image. However, the color images that are publicly released are a
lot more than just art.
>Maybe each photo ought to have a disclaimer pointing out that the color choices
>are arbitrary and have no real relationship to the frequencies of the
>electro-magnetic spectrum as the human eye would see them?
Usually, the color choices are not arbitrary. Assuming the image was made from
frames collected near R, G, and B bands, I think it can be considered to be a
fair representation. I have seen many disclaimers on the color information
released with the original images, although I think those disclaimers sometimes
go astray before the images make it to the popular press.
Jon Isaacs
>from frames collected near R, G, and B bands, I think it can be considered to
be a
>fair representation.
This is what you had previously written which I should have included:
>The objects being imaged would not
>show >color (or very little color) to the human eye no matter how close they
were
>viewed from, and the method of constructing "color" images results in an
>entirely different color space than the eye/brain uses, and often includes
>wavelengths where the eye is insensitive.
So it seems to me that here you saya that these color images are construction
from wave lengths where the eye is insensitive and they are constructed in a
different color space than eye/brain uses.
It seems to me that the color space chosen in these cases must be arbitrary, ie
the choice of the person construction the image.
Anyway, I find these photos less impressive when I realize that they are
processed and modified for maximum appeal. Just my own reaction here, others
may agree, others may disagree.
It would interesting to do a poll, informal or formal addressing awareness of
these issues amount viewers of the photos.
jon
jon isaacs
Trane Francks
> It seems to me that the color space chosen in these cases must be arbitrary, ie
> the choice of the person construction the image.
It's really no more arbitrary than the engineers who develop film
chemistry. Their choice of chemistry dictates whether their film
turns out to be "warm" or "cool". And each photographer has
his/her choice of what constitutes a good film. The colour
weighting that one finds appealing is very subjective.
trane
--
//------------------------------------------------------------
// Trane Francks tr...@gol.com Tokyo, Japan
// Practice random kindness and senseless acts of beauty.
// http://mp3.com/trane_francks/
Lew
> I have seen many disclaimers on the color information
> released with the original images, although I think those disclaimers
> sometimes go astray before the images make it to the popular press.
I think this is the crux of my concern, not that the objects aren't
depicted as the human eye would see them, but that the alterations --
and reasons for the alterations -- are not adequately explained to the
viewer.
Lew
Chris L Peterson
>So it seems to me that here you saya that these color images are construction
>from wave lengths where the eye is insensitive and they are constructed in a
>different color space than eye/brain uses.
>
>It seems to me that the color space chosen in these cases must be arbitrary, ie
>the choice of the person construction the image.
Many of the images are made with the intent of producing "realistic" color
images, and those are made by traditional tri-color imaging, using R, G, and B
filters. The resulting three images are viewed using the R, G, and B channels of
the display device, and the final result is a reasonable approximation of how
the object might appear if it were much brighter, and could stimulate human
color vision. But the color spaces of the camera, display, and eye are all
different, so there are many opportunities for things to go a bit wrong. The
person preparing the image needs to play around with the color balance to try
and get something closer to how he imagines it should look. He may also increase
the saturation to emphasize areas of hydrogen or oxygen. Is this cheating in
some way? I don't think so, but I can see how it might diminish the appeal of
the image for some people.
BTW, virtually all astroimages, whether from earthbound instruments or the HST,
are contrast stretched, usually in a non-linear fashion. This is another "cheat"
that needs to be applied to deal with non-linearities and range limitations of
both our eyes and the display device.
If you were able to travel out to one of these nebulous objects, what you would
actually see in most cases would be a dim gray glow covering a large area, with
faint surface detail- pretty much what you see at the eyepiece of a telescope!
Jon Isaacs
>arbitrary, ie
>> the choice of the person construction the image.
>
>It's really no more arbitrary than the engineers who develop film
>chemistry. Their choice of chemistry dictates whether their film
>turns out to be "warm" or "cool".
>And each photographer has
>his/her choice of what constitutes a good film. The colour
>weighting that one finds appealing is very subjective.
This certainly one level of adjustment.
My concern however is when frequencies invisable to the human eye are mapped
into colors that are visable to the human eye. Not that there is anything
wrong with this, no more than when say a stress measurement is mapped into
color space.
jon
Jon Isaacs
>images, and those are made by traditional tri-color imaging, using R, G, and
>B
>filters.
>The resulting three images are viewed using the R, G, and B channels of
>the display device, and the final result is a reasonable approximation of how
>the object might appear if it were much brighter, and could stimulate human
>color vision. But the color spaces of the camera, display, and eye are all
>different, so there are many opportunities for things
> I don't think so, but I can see how it might diminish the appeal of
>the image for some people.
Small adjustments such as you discuss here are not a problem for me, I think
that is fine.
Its when those "invisable" frequencies are mapped into visable color space that
raises my eyebrows.
jon
Chris L Peterson
>Its when those "invisable" frequencies are mapped into visable color space that
>raises my eyebrows.
Well, I just think of it as a realistic image of the object as it would appear
if my eyes had a wider range of sensitivity <g>. Not unlike slowing down a
recording to hear bats, or speeding one up to hear infrasound.
These pseudocolor images don't bother me, but I think it is important that the
details of processing be included. In fact, those details are released with the
pictures, but when the images show up in non-scientific publications that
information is often left out. That is unfortunate, and does the public a
disservice by (unintentionally) misleading them about the actual appearance of
some very interesting objects.
Chris L Peterson
>I think this is the crux of my concern, not that the objects aren't
>depicted as the human eye would see them, but that the alterations --
>and reasons for the alterations -- are not adequately explained to the
>viewer.
Agreed. The processing details are generally available when the images are
released, but the popular (non-scientific) press seldom passes that information
on. So many people end up not really understanding what they are seeing, which
is too bad.
Doug Goncz
>The objects being imaged would not show
>color (or very little color) to the human eye no matter how close they were
>viewed from, and the method of constructing "color" images results in an
>entirely different color space than the eye/brain uses
Yes. Small bright objects can have a small color, while small low-contrast
objects show a greatly distorted color to the eye.
NTSC color TV takes this into account somewhat, with its monochrome-compatible
Y component, and lowpass filtered U and V on subcarrier. One of U and V is more
or less skin tones, and the other, IIRC, is the nearly invisible green-magneta
axis.
I remember on my Commodore 64 and on CGA dsplays, magenta-green (either way)
text was just horrible, but you could look real close and see the reproduction
was accurate. The two components, at white scale, are the same luminance. But
even this changes on a dim screen, with the Purkinje effect.
On a moonless night, away from the city, blue LEDs make a great flashlight. The
eye is quite sensitive to blue in a dark environment, with dark adaptation
specifically.
A 50X light gathering objective with a 50X magnification eyepiece and a subject
distance of 50 kiloparsecs would produce the same impression as a direct view
from 1 parsec, if unenhanced. Yes?
But to put this on paper, well right there, you've got CMYK color space, a four
dimensional color space, not at all like the three dimenions of vision, whether
you call them RGB or HSV or YUV. They call it the trivariance of vision. The
print gamut is about a third or less of the area of the RGB gamut, which is
itself about 90% of the visual gamut, at any given intensity.
That's why folks used to go for slide shows. If you drink enough malt liquor,
the colors themselves become the show. Whee!
Yours,
Doug Goncz, Replikon Research, Seven Corners, VA
http://users.aol.com/DGoncz
If a computer won't do what needs to be done, lie to it.
Don't try this trick on people.
Jon Isaacs
>appear
>if my eyes had a wider range of sensitivity <g>. Not unlike slowing down a
>recording to hear bats, or speeding one up to hear infrasound.
A good analogy assuming that the frequencies are correspondingly mapped so that
UV doesn't end up red and IR green....
>These pseudocolor images don't bother me, but I think it is important that
>the
>details of processing be included.
This seems to be the concensus, even I agree... <g>
jon
Chris L Peterson
>A good analogy assuming that the frequencies are correspondingly mapped so that
>UV doesn't end up red and IR green....
While I'm sure there are exceptions, most of the out-of-visual-band images I've
encountered map the longer wavelengths to red, and the shorter ones to blue. In
essence, the actual wavelength range is being mapped to the eye's response in
much the same way that the brightness values are stretched to accommodate the
dynamic range of displays.
Sometimes it makes sense to apply more unusual color mappings to emphasize
certain structures, or to show a relationship between different physical
processes going on. While this may result in a beautiful image, and a
scientifically valuable one, any (chromatic) resemblance to the actual object is
pretty clearly lost.