Adobe Illustrator - Levels vs Curves?

Author

Levels vs Curves?

KatWoman

2005-05-26, 7:14 pm

Does it matter which type of adjustment?

To correct exposures I find it easier to use levels. Assuming I don't move
the far left and right points (which I have been told "clips" info out of
your document) is there any reason why using curves is superior?
Also I always use this on an adjustment layer so if it's not "correct" it
can be changed. I like seeing the histogram when making adjustments.
I used to like using the curve palette in my scanner software but the PS one
doesn't work similarly, it tends to make little bumpy mountains instead of a
smooth overall curve, perhaps I am not using the tool well.

tacit

2005-05-26, 7:14 pm

In article <p2ole.20890$lQ3.6922@bignews5.bellsouth.net>,
"KatWoman" <JoliePrincessKatana@hotmail.com> wrote:

> To correct exposures I find it easier to use levels. Assuming I don't move
> the far left and right points (which I have been told "clips" info out of
> your document) is there any reason why using curves is superior?

Moving the far left and right points only clips image detail if you move
them in past the beginning or end of the histogram.

If you have an image whose histogram is, say, all on the left-hand
side--there are no pixels on the right-hand side, meaning there are no
hilight pixels--then moving the right-hand end point to the beginning of
the histogram does not clip detail.

When you move the center triangle, what you are doing is identical to
clicking on the very center in the Curves command and moving it up or
down. The Curves command allows oyu to do far more, however. It allows
you, for example, to increase or decrease contrast, to lighten only the
shadows without affecting the hilights, to lighten only the midtones
without affecting the hilights or shadows, to increase contrast only in
hilights, and so on, and so on. None of these can be done with Levels.

--
Art, photography, shareware, polyamory, literature, kink:
all at http://www.xeromag.com/franklin.html

Bill Hilton

2005-05-26, 7:14 pm

>KatWoman writes ...
>
>Does it matter which type of adjustment?
>To correct exposures I find it easier to use levels.

Levels lets you move three points, the two ends and the middle. If
that's all you need, it's fine, especially since you see the histogram,
which makes it easier to set end points without clipping.

Curves lets you move any of 256 points so it's much more flexible,
especially in the quarter and three quarter tonal areas which you can't
reach directly with Levels. One thing you can do easily with Curves
but you can't do with Levels, for example, is apply an S-curve to
increase contrast (or reverse S to decrease contrast). I also find it
easier to remove a color cast with Curves than with the other tools
since it's easy to click on a spot to set a point on the curve and then
modify it with the arrow keys.

Curves is not as intuitive but I personally think it's well worth the
effort to learn how to use it. Many scanning programs have Curves
superimposed over a histogram, often with the equivalent of Levels at
the bottom of the screen, so you can make an adjustment with either
tool and see it in the histogram. I wish Photoshop did this (at least
the histogram part), it would make using Curves more intuitive.

Bill

Bart van der Wolf

2005-05-26, 7:14 pm

"Bill Hilton" <bhilton665@aol.com> wrote in message
news:1117139642.851871.235510@g14g2000cwa.googlegroups.com...
SNIP
> Many scanning programs have Curves superimposed over
> a histogram, often with the equivalent of Levels at the
> bottom of the screen, so you can make an adjustment with
> either tool and see it in the histogram. I wish Photoshop
> did this (at least the histogram part), it would make using
> Curves more intuitive.

In a way (in a seperate display), Photoshop CS and CS2 do just that
with the life update of the histogram which, unlike the Levels
histogram that only shows R+G+B quantities per level, allows a
multitude of histogram representations (amongst others luminocity
which is closer to our perception of brightness levels).

Bart

Hecate

2005-05-26, 7:14 pm

On Thu, 26 May 2005 14:00:05 -0400, "KatWoman"
<JoliePrincessKatana@hotmail.com> wrote:

>Does it matter which type of adjustment?
>
>To correct exposures I find it easier to use levels. Assuming I don't move
>the far left and right points (which I have been told "clips" info out of
>your document) is there any reason why using curves is superior?
>Also I always use this on an adjustment layer so if it's not "correct" it
>can be changed. I like seeing the histogram when making adjustments.
>I used to like using the curve palette in my scanner software but the PS one
>doesn't work similarly, it tends to make little bumpy mountains instead of a
>smooth overall curve, perhaps I am not using the tool well.
>
You're limited to three points in levels. You have 256 points to play
with in curves. I know it's Photoshop, but a good way of seeing why
this is important is to read Dan Margulis' Professional Photoshop.

--

Hecate - The Real One
Hecate@newsguy.com
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

KatWoman

2005-05-26, 11:15 pm

Thanks, I just tried using the curves and did get much better results in
overall color than with levels.
I wasn't using the tool correctly, it wasn't bending where I wanted but now
I get it.

"tacit" <tacitr@aol.com> wrote in message
news:tacitr-460F89.14185226052005@news-server2.tampabay.rr.com...
> In article <p2ole.20890$lQ3.6922@bignews5.bellsouth.net>,
> "KatWoman" <JoliePrincessKatana@hotmail.com> wrote:
>
>
> Moving the far left and right points only clips image detail if you move
> them in past the beginning or end of the histogram.
>
> If you have an image whose histogram is, say, all on the left-hand
> side--there are no pixels on the right-hand side, meaning there are no
> hilight pixels--then moving the right-hand end point to the beginning of
> the histogram does not clip detail.
>
> When you move the center triangle, what you are doing is identical to
> clicking on the very center in the Curves command and moving it up or
> down. The Curves command allows oyu to do far more, however. It allows
> you, for example, to increase or decrease contrast, to lighten only the
> shadows without affecting the hilights, to lighten only the midtones
> without affecting the hilights or shadows, to increase contrast only in
> hilights, and so on, and so on. None of these can be done with Levels.
>
> --
> Art, photography, shareware, polyamory, literature, kink:
> all at http://www.xeromag.com/franklin.html

hearsay@att.net

2005-05-27, 7:16 pm

KatWoman wrote:
>
> Does it matter which type of adjustment?
>
> To correct exposures I find it easier to use levels. Assuming I don't move
> the far left and right points (which I have been told "clips" info out of
> your document) is there any reason why using curves is superior?
> Also I always use this on an adjustment layer so if it's not "correct" it
> can be changed. I like seeing the histogram when making adjustments.
> I used to like using the curve palette in my scanner software but the PS one
> doesn't work similarly, it tends to make little bumpy mountains instead of a
> smooth overall curve, perhaps I am not using the tool well.

Bill has already provided an excellent explanation between levels and
curves. Jay Arraich went one step further with his detailed article
here, although its demonstration is based only on a single b/w image.

http://www.arraich.com/ps6_tips_ccontrast1.htm

Like Mike and others, I cut my teeth on curves after reading Margulis
(caution: NOT for beginners, VERY opinionated, click by click tutorials
are NON-existent, you better LIKE his writing style, etc.). His theories
are excellent and made me a convert, even though I still don't
understand everything he says. For a period, I tried to rely only on
curves for all my corrections. While curves are as powerful as others
mentioned, I also find problems with them, such as:

- It is tedious to use curves well. For critical corrections (and if not
critical, why bother with curves to begin with?), I have to know which
channel to move, where on each channel to move, how much to move, how
many channels to move, etc. When a correction requires multiple point
moves, your curves window can stay open for quite a while. And that's
assuming that you have analyzed the problem correctly BEFORE moving
curves, and that you DON'T change your mind later. Unwinding all (or
some of) your curve corrections and starting over again is a REAL drag.
Pardon the pun.

- While curves let you increase contrast for one part of the image, it
is at the expense of decreasing contrast somewhere else. Sometimes the
tradeoff is not easy or desirable.

- Correcting for contrast with curves may screw up the colors, and vice
versa.

- As Margulis points out, sometimes correcting with curves within a
single color space just won't do the trick. Borrowing and blending
channels from other color spaces is the solution. To do that, you better
be able to walk on water.

- The power of curves can result in excessive data destruction (8 or 16
bits), and unnatural looking images. OK, same holds true for other
tools.

(Now, why don't $50 books tell you these things?)

I still use curves, but now also rely on other tools such as Selective
Colors (which I don't think curves can do). Jay Arraich goes as far as
calling curves "abominable":

http://www.arraich.com/ps8_CurvesCommentary1.htm

Robert Feinman

2005-05-27, 7:16 pm

In article <4297134A.2EFA1AB1@att.net>, hearsay@att.net says...
> I still use curves, but now also rely on other tools such as Selective
> Colors (which I don't think curves can do). Jay Arraich goes as far as
> calling curves "abominable":
>
> http://www.arraich.com/ps8_CurvesCommentary1.htm
>
>
This commentary is just foolish. In conventional photography
there is a "curve" applied by the response of the film to light.
There is another "curve" applied by the type and degree of developing
and yet others for the print and print developing.
Even viewing on a screen applies a curve.
The eye's response to light is logarithmic: a curve.
Digital cameras have linear response to light, but this produces an
unnatural effect so the image is processed through a curve.
Proper application of curves is much more powerful that levels. And
everything you do with levels can be done with an equivalent curve.
The only feature missing from the curves dialog is the ability to
preview the clipping by pressing ALT while moving the levels end points.
Even in this case it is possible to note the values and transfer them
to a curve after dismissing the levels dialog.
More in the tips section of my web site.

--
Robert D Feinman
Landscapes, Cityscapes and Panoramic Photographs
http://robertdfeinman.com
mail: robertdfeinman@netscape.net

Mike Russell

2005-05-27, 7:16 pm

hearsay@att.net wrote:
> KatWoman wrote:
>
> Bill has already provided an excellent explanation between levels and
> curves. Jay Arraich went one step further with his detailed article
> here, although its demonstration is based only on a single b/w image.
>
> http://www.arraich.com/ps6_tips_ccontrast1.htm
>
> Like Mike and others, I cut my teeth on curves after reading Margulis
> (caution: NOT for beginners, VERY opinionated, click by click
> tutorials
> are NON-existent, you better LIKE his writing style, etc.). His
> theories
> are excellent and made me a convert, even though I still don't
> understand everything he says. For a period, I tried to rely only on
> curves for all my corrections. While curves are as powerful as others
> mentioned, I also find problems with them, such as:
>
> - It is tedious to use curves well. For critical corrections (and if
> not critical, why bother with curves to begin with?), I have to know
> which channel to move, where on each channel to move, how much to
> move, how many channels to move, etc.

I don't agree that curves are tedious. I think they are fun and I have
managed to convince others of this countless times in my class. I think
you're overstating the complexity, and not really describing how curves are
normally used. Much of the changing and rearrangement of curves is
experimental in nature. It is a process of feedback and compromise based on
how the image looks, and how the numeric color values change. For example,
I will often try using an S curve and an inverted S on an image to see which
looks best.

> When a correction requires multiple point
> moves, your curves window can stay open for quite a while. And that's
> assuming that you have analyzed the problem correctly BEFORE moving
> curves, and that you DON'T change your mind later. Unwinding all (or
> some of) your curve corrections and starting over again is a REAL
> drag. Pardon the pun.

But these objections are the same for any color correction method, whether
it involves curves or not. Although it helps, there is no need to predict
anything before using curves, and Photoshop's history mechanism makes
undoing of previous curve operations simple. Starting over again is as
simple as ctrl-Z, and you may use history, adjustment layers or re-load a
curve file if you want to modify your previous curves.

> - While curves let you increase contrast for one part of the image, it
> is at the expense of decreasing contrast somewhere else. Sometimes the
> tradeoff is not easy or desirable.

But the whole goal of using curves is to perform exactly this kind of
trade-off: adding color and contrast to some areas of an image, while taking
it away from others.

> - Correcting for contrast with curves may screw up the colors, and
> vice versa.

This is a true, and often ignored, fact about the RGB and CMYK master
curves - they can cause unwanted changes to color hues, and should not
really be used. Lab avoids this problem by separating color and brightness
information into the Lightness and the a and b channels. Since changing
color is generally more subtle than changing contrast, as a practical matter
the "vice-versa" portion of your statement is not a concern.

> - As Margulis points out, sometimes correcting with curves within a
> single color space just won't do the trick. Borrowing and blending
> channels from other color spaces is the solution. To do that, you
> better be able to walk on water.

One common way to blend channel data is by use of Selective Color. I do
this all the time and it can often fix an image in a matter of seconds. And
your feet will stay dry.

> - The power of curves can result in excessive data destruction (8 or
> 16 bits), and unnatural looking images. OK, same holds true for other
> tools.

I think you've answered your own question in that any image modificationk,
by its nature, changes data values. I would add that data is not destroyed
by moving it around. If a steeper curve makes the image look better, so be
it.

> (Now, why don't $50 books tell you these things?)
Good point. That's why we have Usenet. :-)

> I still use curves, but now also rely on other tools such as Selective
> Colors (which I don't think curves can do).

Lab comes close, however I would be the last person to say curves are the
only tool available.

> Jay Arraich goes as far as
> calling curves "abominable":
>
> http://www.arraich.com/ps8_CurvesCommentary1.htm

Fun article - I'm always ready to enjoy a little icono-clasm :-)> But it is
inaccurate in several respects, not the least of which is the fact that
Ansel Adams did not rely just on dodging and burning, but spent much time
and effort manipulating the characteristic curve of both film and print by
modifying developer chemistry. In fact, changing paper grades is very much
like using Levels or Contrast - altering developing techniques, as Adams
did, foreshadowed Curves.
---
Mike Russell
www.curvemeister.com

KatWoman

2005-05-27, 7:16 pm

Thank you all for the in-depth replies. I used to always use curves in my
scanner software and after comparison of levels vs. curves in a particular
underexposed frame, I find the curve adjustment to be far superior. As I
usually like to pull more from the shadow and less from the highlight,
pulling the curve at a point above the center is better for me. Of course it
would be great to have a perfect in-camera exposure, but here in sunny
Florida the big puffy clouds going overhead caused a great variety of
exposures (similar to bracketing) and wouldn't you know it the client liked
one of the dark frames best. (We did try and pause for the clouds but still
snapped off a few frames before waiting for them to pass.)
I have also used selective color to further adjust the magenta sweater she
wore as it comes out reddish on digital. Even with film I find purple and
magenta often don't register enough blue and look reddish instead of
pinkish.

I am also experimenting with a new sharpen technique I saw on the NG but
have modified to my own technique.
I am making a dupe layer of my orig, put it under orig layer, oversharpen
the under layer with unsharp mask, put original on top on COLOR blending
mode, then erase original on the skin only, keeping it softer focus and
allowing the sharpening to show through the rest. So now everything is
sharper except on the skin where I don't want it sharp. I tried using the
high pass filter as the under layer but wasn't happy with my results.
And yes I do see that curves can make your whole image look crazy colors, in
fact I have used this technique to make some wild looking images. Now I will
use curves to make corrections as well.

Now I tried the little pencil thing in the tool and it makes breaks in the
curve line but I am not sure how or when I would use that. Thankfully my
photographer gives me properly exposed and focused originals most of the
time.

"Mike Russell" <REgeigyMOVE@pacbellTHIS.net> wrote in message
news:6fGle.21785$J12.17541@newssvr14.news.prodigy.com...
> hearsay@att.net wrote:
>
> I don't agree that curves are tedious. I think they are fun and I have
> managed to convince others of this countless times in my class. I think
> you're overstating the complexity, and not really describing how curves
> are
> normally used. Much of the changing and rearrangement of curves is
> experimental in nature. It is a process of feedback and compromise based
> on
> how the image looks, and how the numeric color values change. For
> example,
> I will often try using an S curve and an inverted S on an image to see
> which
> looks best.
>
>
> But these objections are the same for any color correction method, whether
> it involves curves or not. Although it helps, there is no need to predict
> anything before using curves, and Photoshop's history mechanism makes
> undoing of previous curve operations simple. Starting over again is as
> simple as ctrl-Z, and you may use history, adjustment layers or re-load a
> curve file if you want to modify your previous curves.
>
>
> But the whole goal of using curves is to perform exactly this kind of
> trade-off: adding color and contrast to some areas of an image, while
> taking
> it away from others.
>
>
> This is a true, and often ignored, fact about the RGB and CMYK master
> curves - they can cause unwanted changes to color hues, and should not
> really be used. Lab avoids this problem by separating color and
> brightness
> information into the Lightness and the a and b channels. Since changing
> color is generally more subtle than changing contrast, as a practical
> matter
> the "vice-versa" portion of your statement is not a concern.
>
>
> One common way to blend channel data is by use of Selective Color. I do
> this all the time and it can often fix an image in a matter of seconds.
> And
> your feet will stay dry.
>
>
> I think you've answered your own question in that any image modificationk,
> by its nature, changes data values. I would add that data is not
> destroyed
> by moving it around. If a steeper curve makes the image look better, so
> be
> it.
>
> Good point. That's why we have Usenet. :-)
>
>
> Lab comes close, however I would be the last person to say curves are the
> only tool available.
>
>
> Fun article - I'm always ready to enjoy a little icono-clasm :-)> But it
> is
> inaccurate in several respects, not the least of which is the fact that
> Ansel Adams did not rely just on dodging and burning, but spent much time
> and effort manipulating the characteristic curve of both film and print by
> modifying developer chemistry. In fact, changing paper grades is very
> much
> like using Levels or Contrast - altering developing techniques, as Adams
> did, foreshadowed Curves.
> ---
> Mike Russell
> www.curvemeister.com
>
>

KatWoman

2005-05-27, 7:16 pm

I thought I cross posted to the PS groups, sorry to you Illy people I didn't
mean to put this into your NG.

"KatWoman" <JoliePrincessKatana@hotmail.com> wrote in message
news:p2ole.20890$lQ3.6922@bignews5.bellsouth.net...
> Does it matter which type of adjustment?
>
> To correct exposures I find it easier to use levels. Assuming I don't move
> the far left and right points (which I have been told "clips" info out of
> your document) is there any reason why using curves is superior?
> Also I always use this on an adjustment layer so if it's not "correct" it
> can be changed. I like seeing the histogram when making adjustments.
> I used to like using the curve palette in my scanner software but the PS
> one doesn't work similarly, it tends to make little bumpy mountains
> instead of a smooth overall curve, perhaps I am not using the tool well.
>
>

steggy

2005-05-27, 7:16 pm

KatWoman wrote:
>
> I thought I cross posted to the PS groups, sorry to you Illy people I didn't
> mean to put this into your NG.

Well hell, you did now just to make sure!!

LOL no worries

--
steg

Jim

2005-05-27, 7:16 pm

"KatWoman" <JoliePrincessKatana@hotmail.com> wrote in message
news:p2ole.20890$lQ3.6922@bignews5.bellsouth.net...
> Does it matter which type of adjustment?
Yes. Sometimes one is better; sometimes the other. It all depends on what
you need to do to the image.
Jim

tacit

2005-05-30, 7:31 am