color casts and neutral points
Denis Perelyubskiy
I found this tutorial for removing color casts:
http://www.dpchallenge.com/tutorial.php?TUTORIAL_ID=24
Basically, just teaches one to set white/black/neutral points on the
curves dialog.
Neutral gray point is, supposedly, 128%/128% total ink.
What is the significance of 128% ?
What happens when I can't quite find this "perfect" point?
In another tutorial I've seen something about picking a color which
contains equal R/G/B values, which is supposedly also a neutral gray. Is
128% total ink precisely that? (btw, this "perfect" RGB point is also
quite elusive)
Sorry, these seem to do more with color management than with photoshop,
I suppose.
thanks
ps. For example, I am trying to edit the following pictures:
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_0325.html
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_0329.html
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_331.html
These are _after_ some color-cast removal. I am beginning to think that
the first one is not color-cast at all. Also, in the first/second
pictures I can't find that @$%@^ 128% point!
Mike Russell
tell you where the neutral points are - you need to know, or at least
suspect, that an object in the picture is neutral, and then use that
knowledge to remove the color cast.
Here's another tutorial that uses Curvemeister, that explains the concept of
shadow, highlight, and neutral.
http://www.curvemeister.com/tutorials/index.htm
http://www.curvemeister.com/tutorials/Video/Shadow-Highlight-Neutral.avi
The following might be helpful, from the monthly curvemeister curves class:
A neutral point is an object or area in the image that you positively
know, from your everyday experience, has no color. Examples would be
concrete sidewalks, roadways, white clothing, bottoms of clouds, and
other such objects. Usually you will want to find a neutral point
that is near the horizontal midpoint of the curve, but this is not
necessary.
> ps. For example, I am trying to edit the following pictures:
>
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_0325.html
>
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_0329.html
>
http://public.fotki.com/theukrainian/tripstravel/india_bangalore/dsc_331.html
These are interesting images. In each case, find something in the image
that you think should be white, that isn't, and set an eyedropper sample
point there. Adjust curves to make the RGB values for that point equal, and
you're golden. Do this also in Lab, and set a and b to zero at that
oint - you may like that result even better. Then set the shadow and
highlight points, if they exist. I use these terms, as opposed to black and
white points, advisedly, because they are neutral points in their own right.
> These are _after_ some color-cast removal. I am beginning to think
> that the first one is not color-cast at all. Also, in the first/second
> pictures I can't find that @$%@^ 128% point!
The first one does not have much of a color cast. The shadows need to be
opened up a bit, and it wouldn't hurt to bump the saturation a little. Keep
at it, you'll get there. And you may want to get on the list for the
curvemeister class, which is free and does not require that you own
curvemeister.
--
Mike Russell
www.curvemeister.com
www.geigy.2y.net
cb
You may just want to use "Auto Color" in PS, it tries to find grays
and makes them neutral in cast (but not always successful).
The 128 number refers to middle gray. (Based on half of 256, which is
the number of values in each R,G and B channel. It is not a percentage)
You may have the mid-gray eyedropper in Curves set for this, but it
actually ignores the 128 value, just sets neutral values for supposed
neutral pixel in you pictures. I'd get a good book on PS instead of
on-line tutorials. I just purchased PS One-on-One by McClelland. Very
good for novice-advanced.
J Warren
newsg...@dslextreme.com says...
> Couple of questions about why the tutorial I found works.
>
> What happens when I can't quite find this "perfect" point?
I read a cute trick for finding a neutral point to eliminate a color
cast:
1. Create a new layer on top of your image.
2. Set its blending mode to Difference.
3. Use Edit > Fill and pick "50% gray" in the dropdown.
4. Using the Color Sampler tool, look around on the image
until you find an area that appears to be black. Click
the Sampler there. Do this a few more times on other
locations in the image.
5. Delete the 50% Gray layer.
6. Open Levels or Curves and select the Mid-point eyedropper.
7. Click the Color Sampler targets you made in step 4. You
will probably find that one of them causes the color cast
to vanish. Fine tune with Curves if necessary.
Jason
Mike Russell
Although this technique is quite cute, it is more complex and less effective
than simply looking at the image for an object in the image that you know
should be gray, and adjusting curves until you have a neutral there
Chris...@att.net
Looking for a gray object in an image is far from simple. In many
images, there may not be an object that is a true gray (or white, or
black). Knowing how to handle such images is what separates the men from
the boys, or women from the girls. Too bad all the books and tutorials
don't touch on it.
Mike Russell
[re finding the neutral point, if it exists]
> Looking for a gray object in an image is far from simple. In many
> images, there may not be an object that is a true gray (or white, or
> black). Knowing how to handle such images is what separates the men
> from
> the boys, or women from the girls. Too bad all the books and tutorials
> don't touch on it.
Yes, some images - a third or so - don't have a neutral. But when it's
there, t's important to "know" where the neutral is, and not rely on data
from the image to determine it. A similar misconception shrouds the
selection of shadow and highlight, and many people rely simply on the
histogram to find them.
BTW, you're also correct that very little coverage is given to this. Dan
Margulis devotes quite a lot of space to locating neutrals in his
Professional Photoshop book, and Fraser and Haynes at least touch on it in
their respective books.
Your broader point, though, about "adults vs children" touches on another
interesting topic. How do we adjust other important colors, such as flesh
tones, in an image, particularly when no neutral is available in the image?
I'm currently working on support for "color pinning" in Curvemeister, that I
think will take a bite out of that apple, effectively allowing flesh tones,
foliage, blue sky, and other natural colors, to play the role of a neutral
when none is present, or when additional correction to important colors -
catalog matching or logo colors for example - is required.
patrick
the simplest with just a few iterations. Note that it also identifies
neutral areas at varying tone levels for still greater effect.
Combine it with the use of a threshold layer to find and monitor the lo/hi
points and you have a nice start for your curve layer.
. . . . patrick
<Chris...@att.net> wrote in message news:41F7AF8E...@att.net...
Mike Russell
[re using a 50% gray difference layer to locate neutrals]
> I emphatically agree with Chris. This is a very direct technique and
> becomes the simplest with just a few iterations. Note that it also
> identifies neutral areas at varying tone levels for still greater
> effect.
I'm intrigued then. If you are claiming that using a difference layer to
locate values that are "almost" neutral is helpful, I question the value of
that, compared with finding something that you know should be neutral, such
as a bride's dress, newspaper, sidewalk, or certain areas of clouds..
OTOH, it sounds as if you are very convinced that this works, and it does at
least have the feature that it references the image in some way. Can you
illustrate with an example, or describe the technique some more?
> Combine it with the use of a threshold layer to find and monitor the
> lo/hi points and you have a nice start for your curve layer.
> . . . . patrick
Using threshold is indeed an excellent technique for locating the highlight
and shadow.
Odysseus
J Warren <nob...@invalid.urp> wrote:
> In article <10u8ir6...@corp.supernews.com>,
> newsg...@dslextreme.com says...
> > Couple of questions about why the tutorial I found works.
> >
> ...
>
> > What happens when I can't quite find this "perfect" point?
>
> I read a cute trick for finding a neutral point to eliminate a color
> cast:
>
> 1. Create a new layer on top of your image.
> 2. Set its blending mode to Difference.
> 3. Use Edit > Fill and pick "50% gray" in the dropdown.
That will find areas that appear neutral in the *uncorrected* image, but
it won't tell you what *should* be neutral. Supposing you're getting a
red colour cast, an object that was originally aqua or teal might well
look grey, but that doesn't mean it can be used for calibrating the rest
of the image! And if it already appears perfectly neutral, any automatic
adjustment based on those samples will do nothing at all.
<snip>
--
Odysseus
Ron
"Mike Russell" <REgei...@pacbellTHIS.net> wrote in message
news:FfPJd.6088$8Z1....@newssvr14.news.prodigy.com...
You mean like Pictolabs's iCorrect? Have you seen that? They use the Skin,
Foliage, and Blue Sky neutrals you are talking about.
Ron
Ron
"Ron" <rkr...@rcn.com> wrote in message
news:Y_KdnQXtntm...@rcn.net...
Sorry that's Pictographics iCorrect Editlab 4.01. While you're at it,
check out PhotoTune's 20/20 Color MD and their Skin Tune.
patrick
diplomatic exception I have ever encountered on the net. Thanks for that!
First the cop-out and then my own reservations as to why it should not work.
The cop-out. . . . it works for me. It is the best and most direct route to
color cast correction that I have tried. So far, it has never failed. And I
have the certitude of numbers as opposed to guesses as to achieve the
practical results. So much for anecdotal evidence. ("There you go again!")
My queasy feeling that it is wrong and what wakes me up at night is the fact
that the Difference layer is displaying pixels that it *knows* are identical
to the 50% gray fill as pure black. BUT it "knows" under distorting
lighting. Given that there is a problem with the ambient lighting, it *must*
misread and falsely report what it thinks is gray.
So the logic is all wrong.
Howsomever . . . . let's get pragmatic here. Try this:
Use a threshold layer to mark the hi/low points.
Create a new layer, fill it with 50% gray and set the blend to Difference.
Any pixels that match the 50% fill will be displayed as black.
Mark at least two "gray" spots. I look for spots whose RGB values lie within
a range of 10 levels..
Go to you curve layer and use the eyedroppers on all four points.
I always end up with the best color cast correction I can get -- ready for
tweaking for effect.
The exception I take to the altenative -- "finding something that you know
should be neutral" -- is that, unless you photographed a gray card or
MacBeth chart under the same lighting, you're making a judgment as to what
"should be neutral." Lots of varieties of pavement out there, lots of cloud
bases, lots of old gray mares. (Back in the days of yesteryear with light
meters bigger than today's GPS units, I missed a lot of shots by trusting
that "grass is a good approximation to 18% gray.)
I stumbled on this with the intent of averaging the black pixels reported by
the Difference layer at varying tone levels. That makes better logic but too
old to argue with success..
Again, Mike, thanks for the inquiring (as opposed to disparaging) tone of
your response.
cu . . . . patrick
"Mike Russell" <REgei...@pacbellTHIS.net> wrote in message
news:p9RJd.6134$8Z1....@newssvr14.news.prodigy.com...
REMOVE_THIS@hotmail.com Brian K
Brian
Mike Russell
Thanks for your explanation! I'm intrigued enough to give it a try. It may
well be that flagging the near gray values in this way can be of value in
finding the neutrals in an image. If that's the case, and it may well be,
then I've learned something valuable today!
Mike
Chris...@att.net
Mike Russell wrote:
>
> Chris...@att.net wrote:
>
> [re finding the neutral point, if it exists]
>
> > Looking for a gray object in an image is far from simple. In many
> > images, there may not be an object that is a true gray (or white, or
> > black). Knowing how to handle such images is what separates the men
> > from
> > the boys, or women from the girls. Too bad all the books and tutorials
> > don't touch on it.
>
> Yes, some images - a third or so - don't have a neutral. But when it's
> there, t's important to "know" where the neutral is, and not rely on data
> from the image to determine it.
If an image really has a true neutral AND if the cast is UNWANTED, then
removing the cast makes sense and is relatively easy to do. That's what
all the books and tutorials teach you.
Not sure how you arrive at a third of images don't have a neutral. A
landscape photographer shooting at dawn and dusk will certainly disagree
with you. The majority of his images will not have a true neutral. If he
follows the books and tutorials, he would remove the dawn and dusk cast
he worked so hard for.
A similar misconception shrouds the
> selection of shadow and highlight, and many people rely simply on the
> histogram to find them.
Treating shadow and highlight as pure black and white has the same
problem as described above. Many images simply do not have them.
>
> BTW, you're also correct that very little coverage is given to this. Dan
> Margulis devotes quite a lot of space to locating neutrals in his
> Professional Photoshop book, and Fraser and Haynes at least touch on it in
> their respective books.
Margulis is the only one who spends time on this topic, and that section
alone is worth the price of the whole book.
> Your broader point, though, about "adults vs children" touches on another
> interesting topic. How do we adjust other important colors, such as flesh
> tones, in an image, particularly when no neutral is available in the image?
> I'm currently working on support for "color pinning" in Curvemeister, that I
> think will take a bite out of that apple, effectively allowing flesh tones,
> foliage, blue sky, and other natural colors, to play the role of a neutral
> when none is present, or when additional correction to important colors -
> catalog matching or logo colors for example - is required.
Eismann goes into great details on fresh tones, and Margulis also has
some comments. But they are both referring to fresh tones in a "natural"
lighting, i.e. lighting without a cast. We are back to the same problem:
most fresh tones (except those shot in a studio) are under a lighting
with a cast. Without a true neutral in such images, correcting the fresh
tone and preserving a desired cast compounds the problem.
Software attempting to identify neutrals. foliage, sky, fresh tones,
etc. can only assume that these are in a natural lighting without a
cast. Correcting with this kind of software only works for some images,
but not all. Removing a desired cast (sunrise/sunset, candle light) only
gives you a boring image.
In such cases, it would be great if a software can identify an image's
cast and allow an user to remove or adjust it. It would be like trying
out different filters on a camera before taking a shot.
patrick
I suspect my acceptance of the false gray info in the Difference layer is
due to the fact that PS has a better eye than I.
I was working on an image yesterday where the subject was wearing a big,
black Western hat. Sure looked black to me but PS correctly showed it as
having an excess of blue under a cloudy sky. Essentially, it appears that PS
was better at finding a 128-level gray point than I and I stopped at tht
point.
Here's the trail I followed in coming up with an approach that I believe
resolves the apparent conflict of perception vs measurement and I think is
an absolute path to color cast correction.
A true gray sample at any level will have all three RGB levels exactly the
same, provided there is no color cast. Assumig there is a color cast -- say
blue for now -- any true gray sample will have the R and G channels very
close in value while the blue channel will have a higher value.
We can use this idea in setting gray points at *any* level on our curve.
Find and mark sample points that you assume are true gray. Check the Info
palette to verify that the R and G channels are, say, within 5 levels of one
another and the blue channel is somewhat higher. Do this at varying levels
in the background layer. Then create a curve and set the deviant blue
channel to the average of the R and G channels at these points.
Here are some values I got in looking for true gray points: (126/129/138)
(172/175/183).
At the dark end the differences are not as great but the percentages are
about the same: (32/32/35)
When I set the blue channel to the average of the other two channels on the
curve, my color cast was corrected.
(Yes, I marked and corrected the end points first, using a Threshold layer.)
All the Difference layer was doing for me was leading me directly to the
best 128/128/128 approximations in the image. (That's where I got the
128/128/128 sample above.) If we used a 25% (64) gray fill layer, the
Difference blend should take us directly to a 64/64/xx point. Similarily for
a 75% (190) gray fill layer.
There's more to this but I know it gets tedious.
I'd be beholden to be corrected.
cu . . . . patrick
ps: I left the previous thread attached as I presume the topic is of general
interest.
>>>>>>>>>>>>>>>>>>>>
"Mike Russell" <REgei...@pacbellTHIS.net> wrote in message
news:lHZJd.6676$8Z1....@newssvr14.news.prodigy.com...
Don
<REgei...@pacbellTHIS.net> wrote:
>Using threshold is indeed an excellent technique for locating the highlight
>and shadow.
There's a very sneaky "gotcha" when using PS Threshold!
Namely, Threshold does not use absolute RGB values in equal amounts
but combines them using the standard formula; roughly ~60% green, ~30%
red and %10 blue.
This means that Threshold constantly "lies" making it very easy to
overcompensate some RGB components and undercompensate others.
I only noticed this because the bane of my existence has been trying
to remove the Kodachrome cast caused by Nikon scanner's unique light
source (the "Kodachrome mode" in Nikon Scan doesn't go far enough).
And while I'm at it, I follow this "find the gray spot" (G-spot? ;o))
thread with interest.
I always found the generally given advice very unsatisfactory i.e.
"Use something which kinda, sorta, looks like it should, maybe have a
chance of a definite possibility of being gray... Perhaps...".
I would much prefer an objective, mechanical solution to first get the
gray adjusted baseline. Only then should subjective feelings be
applied e.g. sunset, sunrise, etc.
Don.
Mike Russell
>>
>> [re finding the neutral point, if it exists]
>>
>>> Looking for a gray object in an image is far from simple. In many
>>> images, there may not be an object that is a true gray (or white, or
>>> black). Knowing how to handle such images is what separates the men
>>> from
>>> the boys, or women from the girls. Too bad all the books and
>>> tutorials don't touch on it.
Mike:
>> Yes, some images - a third or so - don't have a neutral. But when
>> it's there, t's important to "know" where the neutral is, and not
>> rely on data from the image to determine it.
Chris:
> If an image really has a true neutral AND if the cast is UNWANTED,
> then removing the cast makes sense and is relatively easy to do.
> That's what all the books and tutorials teach you.
In cases where fine tuning of color is required, I and others maintain that
the cast is, by definition, unwanted. Removing a cast is the key to color
correction because it gives the other colors in the image more elbow room,
and opens up the whole image. With a cast in place, all the other colors
are under a veil that is is our job to remove.
I prefer to classify images with "wanted casts", such as sunsets, as special
effects. Such images are easier to color correct because fine tuning of
natural colors is simply out the window.
Chris:
> Not sure how you arrive at a third of images don't have a neutral. A
> landscape photographer shooting at dawn and dusk will certainly
> disagree with you. The majority of his images will not have a true
neutral. If
> he follows the books and tutorials, he would remove the dawn and dusk
> cast he worked so hard for.
An orange sunset is like the Wild West. Anything goes. There is much more
license to use extreme colors and curves this situation, and much less
concern for nuance.
Mike:
>> A similar misconception shrouds the
>> selection of shadow and highlight, and many people rely simply on the
>> histogram to find them.
Chris:
> Treating shadow and highlight as pure black and white has the same
> problem as described above. Many images simply do not have them.
Yes. Shadow and highlight are special cases of neutrals, being the darkest
and brightest areas that contain significant detail, and are neutral. And,
as both of us have said repeatedly, not all images have them.
Mike:
>> BTW, you're also correct that very little coverage is given to this.
>> Dan Margulis devotes quite a lot of space to locating neutrals in his
>> Professional Photoshop book, and Fraser and Haynes at least touch on
>> it in their respective books.
Chris:
> Margulis is the only one who spends time on this topic, and that
> section alone is worth the price of the whole book.
Agreed.
Mike:
>> Your broader point, though, about "adults vs children" touches on
>> another interesting topic. How do we adjust other important colors,
>> such as flesh tones, in an image, particularly when no neutral is
>> available in the image? I'm currently working on support for "color
>> pinning" in Curvemeister, that I think will take a bite out of that
>> apple, effectively allowing flesh tones, foliage, blue sky, and
>> other natural colors, to play the role of a neutral when none is
>> present, or when additional correction to important colors - catalog
>> matching or logo colors for example - is required.
Chris:
> Eismann goes into great details on fresh tones, and Margulis also has
> some comments. But they are both referring to fresh tones in a
> "natural" lighting, i.e. lighting without a cast. We are back to the
> same problem: most fresh tones (except those shot in a studio) are under
a lighting
> with a cast. Without a true neutral in such images, correcting the
> fresh tone and preserving a desired cast compounds the problem.
I don't accept the notion of a "desired cast" for the vast majority of
images. Getting rid of casts is one of the most important goals of color
correction, orange sunsets and other special effects notwithstanding.
Chris:
> Software attempting to identify neutrals. foliage, sky, fresh tones,
> etc. can only assume that these are in a natural lighting without a
> cast. Correcting with this kind of software only works for some
> images, but not all. Removing a desired cast (sunrise/sunset, candle
light)
> only gives you a boring image.
I think I have a handle on a way to give less experience people access
appropriate colors in a more automated way. Whether it will work or not,
time will tell, but at least I think it is a new approach.
Chris:
> In such cases, it would be great if a software can identify an image's
> cast and allow an user to remove or adjust it. It would be like trying
> out different filters on a camera before taking a shot.
Exactly - I don't think we disagreed in the first place.
---
Mike Russell
www.curvemeister.com
www.geigy.2y.net
Hecate
>On Wed, 26 Jan 2005 18:17:57 GMT, "Mike Russell"
><REgei...@pacbellTHIS.net> wrote:
>
>>Using threshold is indeed an excellent technique for locating the highlight
>>and shadow.
>
>There's a very sneaky "gotcha" when using PS Threshold!
>
>Namely, Threshold does not use absolute RGB values in equal amounts
>but combines them using the standard formula; roughly ~60% green, ~30%
>red and %10 blue.
>
That, incidentally, is why you should *never* go :
Image/Mode/Grayscale. The conversion is always 3:6:1 regardless of
the image.
--
Hecate - The Real One
Hec...@newsguy.com
veni, vidi, reliqui
Mike Russell
> On Wed, 26 Jan 2005 18:17:57 GMT, "Mike Russell"
> <REgei...@pacbellTHIS.net> wrote:
>
>> Using threshold is indeed an excellent technique for locating the
>> highlight and shadow.
>
> There's a very sneaky "gotcha" when using PS Threshold!
>
> Namely, Threshold does not use absolute RGB values in equal amounts
> but combines them using the standard formula; roughly ~60% green, ~30%
> red and %10 blue.
>
> This means that Threshold constantly "lies" making it very easy to
> overcompensate some RGB components and undercompensate others.
Well, no. Although there are places in Photoshop where this is the case,
Image>Adjust>Threshold just uses the raw channel values, not the composite
luminance values. Your point is still well taken, though, and it is
important to use the information from the Threshold command as an
approximate guide only, and use the eyedropper tool for precise
measurements.
> I only noticed this because the bane of my existence has been trying
> to remove the Kodachrome cast caused by Nikon scanner's unique light
> source (the "Kodachrome mode" in Nikon Scan doesn't go far enough).
Short answer: Vuescan and an IT8 target from Wolf Faust.
> And while I'm at it, I follow this "find the gray spot" (G-spot? ;o))
> thread with interest.
It's been a civil and informative discussion. I almost think we're in a
Yahoo group and not on Usenet :-)
> I always found the generally given advice very unsatisfactory i.e.
> "Use something which kinda, sorta, looks like it should, maybe have a
> chance of a definite possibility of being gray... Perhaps...".
Yes - there is a lot of vague information out there. Histogram worship, as
Margulis calls it, is rampant, and people generally expect there to be an
automated process that will do it all for them. This works, to a point, but
for really good results human judgement is required.
> I would much prefer an objective, mechanical solution to first get the
> gray adjusted baseline. Only then should subjective feelings be
> applied e.g. sunset, sunrise, etc.
I agree with the object mechanical part - that's the eyedropper tool. It's
one of Photoshop's most powerful features, and the best way to detect a
color cast. Imagine what this sort of tool would have cost in the old days
of chromagenic chemical photography: thousands. Adobe is well aware of
this, and so the eyedropper is absent from Elements.
But I don't agree that you should take a sunset back to neutral before
adding back the orange color. Everyone expects and appreciates that a sunset
is a gorgeous orange color. while you might look for a neutral for effect -
for example to bring out contrasting blue or violet in the upper clouds -
there is no need to go through the door of perfect color balance first. By
the same token, it's also a philosophical mistake to insist that, because of
images such as sunsets, that all images should retain a color cast.
Chris...@att.net
Took me awhile to realize the importance of locating these points
precisely before editing. Nothing is worse than correcting the wrong
colors!
Andrew Morton
> Yes. Shadow and highlight are special cases of neutrals, being the
> darkest and brightest areas that contain significant detail, and are
> neutral. And, as both of us have said repeatedly, not all images
> have them.
How about shadows on a grey/white surface under a blue sky? They're blue.
Admittedly, it is rare for that to be a problem when removing a colour cast.
Andrew
Clyde
Wow! You guys certainly make this hard. I guess I'm spoiled by
CurveMeister. I just click on what I think it the highlight, then drag
that point around until I find the real highlight. Then I do the same
for the shadow; the tool shows me when I've hit it.
The neutral is a tad hard, but not much. I click on that I think is a
neutral. Then I drag that point around until the overall picture has the
color cast that I want. Some of those places aren't as neutral as I
thought. Some of those places will give me fine tuning in color cast.
In the rare case of no real neutral in the picture, I will manually
adjust the colors inside CurveMeister. I always use this tool in LAB
mode and a few bumps on the 2 color curves will usually get a color cast
that I like. Unless I'm doing this, the whole process only take a minute
or so. If I do manual adjustment, I can spend 2 or 3 minutes.
I love the speed and simplicity of this tool. And I'm not connected to
them at all - except as a happy customer.
Clyde
Don
>>There's a very sneaky "gotcha" when using PS Threshold!
>>
>>Namely, Threshold does not use absolute RGB values in equal amounts
>>but combines them using the standard formula; roughly ~60% green, ~30%
>>red and %10 blue.
>>
>That, incidentally, is why you should *never* go :
>Image/Mode/Grayscale. The conversion is always 3:6:1 regardless of
>the image.
Exactly!! After discovering this Threshold bias, I converted the image
to Grayscale in order to avoid it - only to find out that Grayscale
does the same thing!
As someone who scans B&W images in color (for easier editing) I
encountered this very early on.
Don.
Don
<REgei...@pacbellTHIS.net> wrote:
>> This means that Threshold constantly "lies" making it very easy to
>> overcompensate some RGB components and undercompensate others.
>
>Well, no. Although there are places in Photoshop where this is the case,
>Image>Adjust>Threshold just uses the raw channel values, not the composite
>luminance values.
There's a very easy test for this: Go to Image/Histogram/Luminance and
then Threshold. You'll see that the histograms are identical.
Now go to Image/Adjust/Levels/RGB and you'll see a totally different
histogram (these are the actual raw channel values simply added up).
All this is in PS6 and I don't know if it has changed since, but I've
done extensive testing creating Excel spreadsheets with color
information and in my experience Threshold does use luminance values.
I actually "discovered" the 6:3:1 ratio empirically, on my own, before
I ever heard of color luminance perception.
>> I only noticed this because the bane of my existence has been trying
>> to remove the Kodachrome cast caused by Nikon scanner's unique light
>> source (the "Kodachrome mode" in Nikon Scan doesn't go far enough).
>
>Short answer: Vuescan and an IT8 target from Wolf Faust.
Oooh... Dangerous territory... ;o)
I'm known to have a very low opinion of VueScan (pop over to
comp.periphs.scanners).
VueScan is extremely bug ridden and highly unreliable. Wouldn't touch
it with a proverbial 10 foot pole.
>> I always found the generally given advice very unsatisfactory i.e.
>> "Use something which kinda, sorta, looks like it should, maybe have a
>> chance of a definite possibility of being gray... Perhaps...".
>
>Yes - there is a lot of vague information out there. Histogram worship, as
>Margulis calls it, is rampant, and people generally expect there to be an
>automated process that will do it all for them. This works, to a point, but
>for really good results human judgement is required.
From what I've read Margulis seems to go to the other extreme, "gut
feeling worship" to coin a phrase.
For me, the automated aspect is really secondary. What I'm concerned
with is the objective part. Of course, anything objective can also be
formalized and then automated. But it's the objective evaluation I
(for one) am interested in. (Also, please see end of message.)
>But I don't agree that you should take a sunset back to neutral before
>adding back the orange color.
No, that's not what I meant. I was referring to identifying and
removing the non-image cast before making esthetic adjustments. The
key is this:
>By
>the same token, it's also a philosophical mistake to insist that, because of
>images such as sunsets, that all images should retain a color cast.
There are two distinct types of cast: the cast inherent to
(caused/introduced by) the medium (e.g. a scanner's light source) and
the cast inherent to the image subject (e.g. a sunset).
The latter cast may or may not be present and is the domain of
subjective, esthetic decisions.
The former, however, is objective and, as such, "curable" by automatic
means. Indeed, using our imperfect and constantly changing subjective
perception to remove this objective artifact is very counter
productive.
Don.
Loren Amelang
>I control drag the sliders in Levels to locate the highlight and shadow.
Does using the Control modifier do something in versions newer that 6?
I can't see any effect in my version. And Google didn't find any
mention of Control+drag in levels sliders... Except yours.
Loren
Mike Russell
Hi Andrew,
As with any shadow, they will start out blue, and blue snow can be a
pleasant effect, particularly if the snow is the main subject of the image,
or if your main subject is in the sunlight, and the snow is acting as
background. For that kind of subject, I would generally make the main
subject neutral, and chalk up the blue shadows to atmosphere. OTOH, if the
subject of the image, and the snow, both have the blue cast, I recommend
removing the blue cast altogether.
The most interesting case is where there are important elements in the
image, some of which are in direct sunlight, and others in blue shadow. In
such mixed lighting situations, it sometimes works to apply two neutrals in
RGB mode - this works because RGB mixes brightness and color information,
and the objects that are in shade are generally darker, causing the curve
control points for the two neutrals to be well separated. Otherwise you can
use a mask, and set a neutral in each masked area.
Our eyes do the same thing. Try it, you'll like it.
Mike Russell
This is a nifty Mac feature that does not work on Windows.
Mike Russell
> On Fri, 28 Jan 2005 00:16:48 GMT, "Mike Russell"
> <REgei...@pacbellTHIS.net> wrote:
>
>>> This means that Threshold constantly "lies" making it very easy to
>>> overcompensate some RGB components and undercompensate others.
>>
>> Well, no. Although there are places in Photoshop where this is the
>> case, Image>Adjust>Threshold just uses the raw channel values, not
>> the composite luminance values.
>
> There's a very easy test for this: Go to Image/Histogram/Luminance and
> then Threshold. You'll see that the histograms are identical.
Darned if you aren't right, Don. I stand corrected. This doesnn't change
our convivial agreement on the original procedure, which is still to use the
Threshold command as a guide for later measurement with the eyedropper.
...
> From what I've read Margulis seems to go to the other extreme, "gut
> feeling worship" to coin a phrase.
Not at all. Rather than gut feeling, Margulis's techniques rely on careful
use of the eyedropper tool to measure neutrals, flesh tones, etc.
(snipping a bit, for clarity)
> The [cast from scanner light source], however, is objective and, as such,
"curable"
> by automatic means. Indeed, using our imperfect and constantly changing
subjective
> perception to remove this objective artifact is very counter productive.
Indeed, it should be possible to get a good scan by generating a profile
from an IT8 target. The spectral characteristics of the light are an issue,
but so is the absorption spectrum of the dyes in the film base. This is why
it is often better to have an IT8 target for your particular film. It may
be that the profile you are using is based on a different film emulsion than
the one you are scanning and this could explain the persistent "kodachrome"
problem you are facing.
Sorry to hear, BTW, that Vuescan is not doing it for you. I've found Ed
Hamrick to be very responsive to requests, so you may want to contact him
directly re specific problems.
Chris...@att.net
> There are two distinct types of cast: the cast inherent to
> (caused/introduced by) the medium (e.g. a scanner's light source) and
> the cast inherent to the image subject (e.g. a sunset).
>
> The latter cast may or may not be present and is the domain of
> subjective, esthetic decisions.
>
> The former, however, is objective and, as such, "curable" by automatic
> means. Indeed, using our imperfect and constantly changing subjective
> perception to remove this objective artifact is very counter
> productive.
Very well put, Don. Much better than my use of "undesired" or "desired"
casts. The first kind you described is what I would call "undesired",
and most will agree that they should be removed. The decision and
technique is relatively easy.
The second kind is what I would call "desired". The decision and
technique in handling it is less straight forward. This kind of cast can
be present in many images where the lighting is not under control (i.e.
not in a studio), or where there are more than one source of lighting.
One object in an image can receive lighting from a main source (e.g. sun
light), as well as reflected light (light bounced off a green wall),
while other objects in the same image do not. This kind of situation is
very common when you think about it, and dealing with this kind of cast
is not simple.
Chris...@att.net
Oops! I meant to say: In Levels, hold down the Alt key (in PC) and drag
the Input slider's highlight or shadow. A more detailed description (and
a complete PS workflow!) can be found here:
Mike Russell
[re removing color casts]
> Wow! You guys certainly make this hard. I guess I'm spoiled by
> CurveMeister. I just click on what I think it the highlight, then drag
> that point around until I find the real highlight. Then I do the same
> for the shadow; the tool shows me when I've hit it.
>
> The neutral is a tad hard, but not much. I click on that I think is a
> neutral. Then I drag that point around until the overall picture has
> the color cast that I want. Some of those places aren't as neutral as
> I thought. Some of those places will give me fine tuning in color cast.
>
> In the rare case of no real neutral in the picture, I will manually
> adjust the colors inside CurveMeister. I always use this tool in LAB
> mode and a few bumps on the 2 color curves will usually get a color cast
> that I like. Unless I'm doing this, the whole process only take a
> minute or so. If I do manual adjustment, I can spend 2 or 3 minutes.
>
> I love the speed and simplicity of this tool. And I'm not connected to
> them at all - except as a happy customer.
But some people do prefer to do remove casts the hard way! It builds
character, like walking to school uphill both ways in the snow. :-)
Uni
Think you need a one button photo fix gadget? Lord, no! Here is an
example of manual color correction, that can beat ANY one button photo
fix feature!:
http://community.webshots.com/photo/42670526IIeqQz
Not even Ron Lacey (author of PSP-8 Zero To Hero) can come close to my
abilities!:
http://community.webshots.com/photo/118879751QdfdDW
:-)
Uni - The Master Colorcast Remover
Don
<REgei...@pacbellTHIS.net> wrote:
>But some people do prefer to do remove casts the hard way! It builds
>character, like walking to school uphill both ways in the snow. :-)
You left out "barefoot" and "in the dark"... ;o)
I plead to this one. Whenever I'm told how wonderful, easy,
life-changing, automatic, etc... some new feature is, my first
reaction is:
Great!!! Now, how do I turn it off? ;o)
Don.
Don
>The second kind is what I would call "desired". The decision and
>technique in handling it is less straight forward. This kind of cast can
>be present in many images where the lighting is not under control (i.e.
>not in a studio), or where there are more than one source of lighting.
>One object in an image can receive lighting from a main source (e.g. sun
>light), as well as reflected light (light bounced off a green wall),
>while other objects in the same image do not. This kind of situation is
>very common when you think about it, and dealing with this kind of cast
>is not simple.
Indeed! The problem is we have to make subjective decisions and that's
always difficult because our subjective feelings change and are
influenced by our imperfect perception.
That's why I belong to the "raw scan sect" ;o) and always archive a
"digital negative" first. That way, if my requirements or feelings
change in the future, I can always start from scratch.
Don.
Don
<REgei...@pacbellTHIS.net> wrote:
>> There's a very easy test for this: Go to Image/Histogram/Luminance and
>> then Threshold. You'll see that the histograms are identical.
>
>Darned if you aren't right, Don. I stand corrected. This doesnn't change
>our convivial agreement on the original procedure, which is still to use the
>Threshold command as a guide for later measurement with the eyedropper.
(Sorry for the delay...)
Oh, absolutely! The eyedropper is essential. For everything!
It's still a shame, though, that Threshold doesn't offer the option of
Luminance or "superimposed-RGB", for a lack of a better word.
In its absence, I first use Levels to identify the channel which is
closest to the edge and then do a Threshold on that channel alone to
identify the affected area of the picture. Or, in extreme cases,
superimpose 3 individual channel Thresholds.
It's convoluted, but the alternative is even worse i.e., create a
"negative Luminance" image to cancel Threshold's compensation.
>> From what I've read Margulis seems to go to the other extreme, "gut
>> feeling worship" to coin a phrase.
>
>Not at all. Rather than gut feeling, Margulis's techniques rely on careful
>use of the eyedropper tool to measure neutrals, flesh tones, etc.
I only read a single chapter, but in there he boasts about teaching
blind people (well, color blind) to do color correction using this
method. It relies on areas which "should" be gray or which "usually"
are gray. And then in the same breath (and in spite of bracketing
these areas with such relative words) he asserts how infallible this
method is. To me that's a contradiction and too much reliance on these
potentially gray areas.
Don't get me wrong, his advice on finding grays is *very* good - and I
benefited a lot from reading even this single chapter - but it's not
as water-tight or world-shattering as presented.
>Sorry to hear, BTW, that Vuescan is not doing it for you. I've found Ed
>Hamrick to be very responsive to requests, so you may want to contact him
>directly re specific problems.
Even if all the many bugs were fixed, VueScan is just not the type of
program I would ever use.
As for the author, over in comp.periphs.scanners he's been very
self-opinionated and arrogant. When his programming failings were
repeatedly exposed with fact he became petulant, threw a temper
tantrum showering obscenities and, allegedly, left. It was ugly...
Don.
Mike Russell
> On Sat, 29 Jan 2005 01:24:15 GMT, "Mike Russell"
> <REgei...@pacbellTHIS.net> wrote:
>
>>> There's a very easy test for this: Go to Image/Histogram/Luminance
>>> and then Threshold. You'll see that the histograms are identical.
>>
>> Darned if you aren't right, Don. I stand corrected. This doesnn't
>> change our convivial agreement on the original procedure, which is
>> still to use the Threshold command as a guide for later measurement
>> with the eyedropper.
>
> (Sorry for the delay...)
>
> Oh, absolutely! The eyedropper is essential. For everything!
>
> It's still a shame, though, that Threshold doesn't offer the option of
> Luminance or "superimposed-RGB", for a lack of a better word.
Camera raw does this, as does - ahem - Curvemeister.
> In its absence, I first use Levels to identify the channel which is
> closest to the edge and then do a Threshold on that channel alone to
> identify the affected area of the picture. Or, in extreme cases,
> superimpose 3 individual channel Thresholds.
>
> It's convoluted, but the alternative is even worse i.e., create a
> "negative Luminance" image to cancel Threshold's compensation.
Yes, that is convoluted, so to speak. :-)
>>> From what I've read Margulis seems to go to the other extreme, "gut
>>> feeling worship" to coin a phrase.
>>
>> Not at all. Rather than gut feeling, Margulis's techniques rely on
>> careful use of the eyedropper tool to measure neutrals, flesh tones,
>> etc.
>
> I only read a single chapter, but in there he boasts about teaching
> blind people (well, color blind) to do color correction using this
> method. It relies on areas which "should" be gray or which "usually"
> are gray. And then in the same breath (and in spite of bracketing
> these areas with such relative words) he asserts how infallible this
> method is. To me that's a contradiction and too much reliance on these
> potentially gray areas.
Dan does come close to boasting in this chapter, however in his defense I
would say that there are a fair number of people who are color blind, who
would also like to produce reasonable photographs. The numeric nature of
Dan's techniques provides a very good - no infallible - way, to accomplish
this. Dan does provide an example image, an off-white wine label, where the
color blind person comes up with the wrong result, and discusses why this
happened.
> Don't get me wrong, his advice on finding grays is *very* good - and I
> benefited a lot from reading even this single chapter - but it's not
> as water-tight or world-shattering as presented.
Again, not the greatest chapter to start with.
>> Sorry to hear, BTW, that Vuescan is not doing it for you. I've
>> found Ed Hamrick to be very responsive to requests, so you may want
>> to contact him directly re specific problems.
>
> Even if all the many bugs were fixed, VueScan is just not the type of
> program I would ever use.
>
> As for the author, over in comp.periphs.scanners he's been very
> self-opinionated and arrogant. When his programming failings were
> repeatedly exposed with fact he became petulant, threw a temper
> tantrum showering obscenities and, allegedly, left. It was ugly...
I've found him to be generally cooperative and polite, but he does lose it
sometimes. This wasn't when he was arguing with Mac people, was it? :-)
Don
<REgei...@pacbellTHIS.net> wrote:
>> Don't get me wrong, his advice on finding grays is *very* good - and I
>> benefited a lot from reading even this single chapter - but it's not
>> as water-tight or world-shattering as presented.
>
>Again, not the greatest chapter to start with.
That's the only one I read but, even with my doubts about the method,
I still learned a lot about finding grays. So even though I may not
use the method I benefited greatly even from this one chapter.
>> As for the author, over in comp.periphs.scanners he's been very
>> self-opinionated and arrogant. When his programming failings were
>> repeatedly exposed with fact he became petulant, threw a temper
>> tantrum showering obscenities and, allegedly, left. It was ugly...
>
>I've found him to be generally cooperative and polite, but he does lose it
>sometimes. This wasn't when he was arguing with Mac people, was it? :-)
I thought he was a Mac person himself? ;o)
No, it was the Minolta streaking in VueScan which put him over the
edge (this time...). Other people wrote afterwards that he has these
tantrums regularly.
For over a year he was unable to fix the streaks only occurring in
VueScan (Minolta software worked fine) and each time he claimed he did
someone would go: "Ahem... The streaks are still there!". Finally, he
just blew up in frustration hurling obscenities indiscriminately.
Not exactly a good advert for VueScan...
Don.
Clyde
By all accounts, I have more "character" than I need already. Besides,
I've never really been a fan of pain-for-gain. You can call this
efficiency or laziness or "working smarter".
To each his own,
Clyde