Question about curves - DPChallenge Forums

11/11/2003 09:54:06 AM · #1

Paul (GeneralE) has talked about curves a number of times, and, as a consequence, I started trying to figure them out. I've been playing around with them, and their use can do some pretty amazing stuff to pictures, BUT, what's eluding me is, "How do they actually work?" That is, what exactly is it that I'm doing when I'm using curves?

I hope this is not another dumb question, but for me it's always important to know what's behind what I'm doing. I use PaintShop Pro, not PhotoShop, so my language is a little different than Paul's.

If anyone knows the answer, please post. And if it's a dumb question, please ignore. The answer is probably something very simple, and I'll go, "Duh!" once again. Oh well.

Take care all,

Ursula

Gordon

RIM BlackBerry Curve

Canon EF 85mm f/1.8 USM

11/11/2003 09:57:53 AM · #2

The curve defines a mapping, between the original tones in the image and the final tones

So you have the input values on the y axis, and the output values along the x axis (or reversed)

The bottom left is black, the top right is white (depends how you've set it up but this is the default)

All the curve is doing is saying for this level in the input, map it to this level in the output - so you can compress and stretch tones in different areas.

This can also apply to colours by extension to each of the R, G and B channels.

Canon 12x IS USM (S3 IS built-in Lens) -- 36-432mm (35mm EQ)

11/11/2003 10:16:21 AM · #3

What you are doing is mapping one value to another. If you click at any point in the picture, you will see a point on the curve showing input and output values.

Most of the time, the values are rated in percentage points; you have an option whether to display the 0,0 point in the lower-left corner (typical for graphs) or the upper-right (Ohotoshop's idiotic default) by clicking on the grayscale displayed below the graph.

If you grab the top-right point and drag it down, you could (for example) lower the 100% mark to 95%. That means nothing in the image would print darker than 95% of the original value; other points on the graph would be proportionally lower.

Its main use is to expand or contract the tone range in parts of the photo. As a simpke example, take the 25% point on the graph and drag it down to 20%. Take the 75% point and move it up to 80%. The part between the two points is now steeper, and the part of the photo formerly represented by 50 values (75-25) is now represented by 60 values (80-20), and will likely have more contrast and detail. The part below the 25% mark will become flatter in tone and lose detail, since what was formerly covered by a range of 25 values now only has 20. Likewise for the shadows above 75%, which now have only the tones between 80-100% to represent them.

In addition to working like this on the composite image, you can also apply a curve separately to each color channel to control or change colors.

I hope this helps make it clearer. I think the best tutorials I've read on Curves are at Luminous Landscape.

Canon 12x IS USM (S3 IS built-in Lens) -- 36-432mm (35mm EQ)

11/11/2003 10:16:59 AM · #4

Darn slow typing ... at least I think we said the same thing ...

Message edited by author 2003-11-11 10:17:26.

willem

Sony Alpha a6700

Sony E 18-135mm f/3.5-5.6 OSS

11/11/2003 10:24:27 AM · #5

They way I find curves very usefull is in two ways :

1) to increase contrast I first put a dot on the middle of the line so that points stays fixed. Then I pick a point on the line at about 1/4 from the bottom left corner and pull it slightly to the bottom right. As result you now have a slight S shapes curves, which darkens the shadows and lightens the highlights.

2) to lighten a picture without blowing out highlights. I pick the middle point and pull it slightly to the top left.

Usually it needs very slight adjustments.

11/11/2003 10:48:25 AM · #6

OK, thanks everyone, so far so good, BUT,

Is there a different way to figure out (to see) the input and output value of a point on a picture? If I understand correctly what you're saying, in PhotoShop you can click or point to any point and you will see this point reflected on the curve, right? I think that's what been throwing me, it doesn't seem you can do that in PSP, and that's why to me working with curves looks like "guessing".

Ursula

PS - I hadn't checked Luminous Landscape, I'll do so later today.

Canon 12x IS USM (S3 IS built-in Lens) -- 36-432mm (35mm EQ)

11/11/2003 10:58:05 AM · #7

Do you have an "info window" which displays the value of the pixel under the cursor? In Photoshop, whenever you have an adjustment (layer) like Curves active, that will then display the before/after values of the cursor's location.

11/11/2003 11:17:04 AM · #8

As far as I can tell, I get the RGB values in info. Then there's a O: which reads 255 usually, and below that it reads "100%". That is in a normal layer. In an adjustment layer for Curves I get only the RGB values of a particular point. Does that make sense?

Canon 12x IS USM (S3 IS built-in Lens) -- 36-432mm (35mm EQ)

11/11/2003 12:12:23 PM · #9

No, when the Curves adjustment layer is active in Photoshop, the info window would read something like:

R 70/67
G 34/31
B 66/57

The 255 is the 8-bit computer value for the color; each color can have a value of 0-255, however, most people find it easier to think in percentages, even though it's a coarser scale.

Actually, I usually have my info window set to display CMYK values instead of RGB values, since most of my work in Photoshop involves preparation of images for offset print -- I just mentally translate trhe CMYK builds into perceptible colors -- however, the principle is exactly the same. Photoshop lets you display two sets of values -- I also have the Grayscale or composite value displayed, e.g.

K 68/64

Canon EF 17-40mm f/4.0L USM

11/11/2003 12:30:13 PM · #10

Ah, OK, now I know what I'm looking for. Your example (of your info readout) helps. I don't get that info in PSP, but I'll figure out how to get it. So greyscale is K. OK. I'm getting this.

Thank you very much, Paul, Gordon, and Willem.

Unrelated question, you don't have to answer it, but I'm curious - why is CMYK used for offset?

Ursula

Gordon

RIM BlackBerry Curve

Canon EF 85mm f/1.8 USM

11/11/2003 12:38:57 PM · #11

Because printers print in Cyan Yellow Magenta and Black (or 'K'ey)

soup

Canon EOS-30D

11/11/2003 12:40:34 PM · #12

CMYK

soup

11/11/2003 12:44:35 PM · #13

Originally posted by Gordon:

Because printers print in Cyan Yellow Magenta and Black (or 'K'ey)

Duh!!

Thanks, Gordon.