Levels and curves - how too advice

So I've not used levels and curves much in so long that I've half forgotten what little I knew about them before. So I've decided its time to get those two tools back into use again and would welcome some tips/advice/links/refe​rences to articles and methods people use for levels and curves.

I'm aware that a lot of photography is "what looks good" but also that when trusting ones own eyes only its possible to spend ages going back and forth between two or three revisions and be unable to make a choice based purely on "what look good". So some sensible reference points and advice with regard to these two tools would be fantastic to have.

Levels is a more linear tool, and easier to master.

Briefly, pull the right side to the left,. this will make the brights brighter,. then pull the left to the right making the darks darker. Then on most levels tools there is a center line that allows you to adjust the mid-tones, but often it's unneeded.

It essentially a more controlled contrast tool, but like separate bass and treble, it is better than "loudness" which boosted both at once.

Curves is a bit like Level's hopped up on goofy pills. It is, well, much more complex, and takes a lot more feel IMHO. I'll leave it to someone that uses it to explain.

If I'm manipulating photos in PS, I'm usually fine tuning, so I'll use curves. The main thing I'll use levels for is if I've scanned a document, and I want to make the paper parts white and text black. The quick and easy way is to go to your levels and select the white eyedropper. Then look for the most gray looking area of "white" and click the eyedropper on it. PS will automatically set the white point to that gray scale value (and all those gray areas will turn white). You can also zoom into some text, then use the black eyedropper to make the text rich and black.

Curves give you more fine adjustments for adjusting just certain portions of contrast, or a particular color channel. If you're using Photoshop, there's a non-destructive method: you can add either a levels or curves adjustment layer on top of your photo layer. If you don't like the results, then you can just discard and try again. Or you can create a layer mask on the adjustment layer, and paint the mask in certain areas of the photo.

Here's a general guide on how to use curves: https://helpx.adobe.co​m …ng/curves-adjustment.html

Thanks both, I think the issue I have is partly that unless its very simplistic (ergo your example Dave of black text on white paper) I almost become lost in choices. Most natural scenes tend to have many different colours from green grass to blue sky. I've often found that using curves/levels to remove a colour haze over the photo tends to also end up favouring one colour over the others so it can appear "better" than the original, but a better look (fresh eyes) can show that I've slowly turned grass a shade of blue etc...

Of course in theory one can use layermasks but then it would be an insane amount of work per-photo and beyond what I think the tools should need to be used for in general use.

Yeah, one can spend a lot of time going back and forth not really deciding what looks best to them. The other day I had a conversation with a friend that's into mixing music. We talked about how one's perception can be different one day to the next, and how that effects sound or image editing. As for color channels with curves...I first will try to optimize contrast with the RGB channel first. Then I'll first decide if there needs to be a color adjust by going to the color channel, and then make a middle point on the diagonal line and pull or push it (to either add or subtract within all the midtones). That way you're starting globally and seeing if its worth it to do more fine tuning. I also have a background in computer graphics, so I've got a nice graphics tablet (and painting layer masks are second nature to me).

Overread wrote in post #18371793
Of course in theory one can use layermasks but then it would be an insane amount of work per-photo and beyond what I think the tools should need to be used for in general use.

Can you just use a brush? Not sure what software you are using.

I find curves more intuitive and rarely use levels. I guess you need to have an idea of were the tones in your image map and where you want them to go and maybe after years of doing B&W it just seems obvious to me.

As I've not really got a background in art learning more about tonality and how to "read" a photo visually like that might be of benefit. At present its one of those "grab the slider and see what happens" situations for me (which I suspect is the case for many people). However if there's any advice on learning how to read the tonality of an image that would be great to hear (although I suspect its one of those skills where there's a short bit of theory and a lot of practice reading different photos to really get to a point where its easy to do).

Curves layer is what finally got me started down the right path with PS.
Years ago a unnamed company had some dvds by various experts. One of them was Ben Willmore. I love his teaching style and explained in common language how curves works. But at the end of the lesson he said "of course you can just try auto to see what you get"

That is exactly what I do then I tweak from there. Sometimes auto is not close to what I want, sometimes it is perfect but it give you the place to start.
In my mind you cant adjust anything else until the white balance is set so I always start with curves.

I cannot find the disk I bought on his site but here is a link to his site.
Gary
https://learntolovepho​toshop.com/learn-to-love-photoshop/

Levels and curves are both powerful tools for tone (re)mapping. Levels is basically a curve with one point between the black and white ends. Both tools are useful to set the image black and white point by adjusting the ends of the curve or dragging the endpoints on the Levels controls.

Curves is a more versatile tool in that you can alter contrast across the tonal range, targeting specific tonal ranges for more contrast (making a segment of a curve steeper) or less contrast (making a segment of the curve flatter). Just remember that if you want to increase contrast in one tonal range, adjacent tonal ranges will receive less contrast as they get flattened. Take for example the typical "S" curve that is used to add contrast to the midtones, mimicking our visual perception of a scene. The middle part of the "S" steepens the curve in the midtones, adding contrast; however, the shadows and highlights get compressed and may lose detail and contrast. Similarly, when working with RGB curves, adding contrast alters color and saturation, another unintended (perhaps) consequence of remapping tones in the image. One can counter this effect by using a Luminosity blend mode or working in Lab and adjusting the L curve.

Curves are also very useful when individual channels are used to manipulate the color in a specific tonal range. If you shoot in shade and your shadows are too blue, you can use the R, G and B channel curves to alter the shadow white balance to a more neutral or warm tone to match the non-shadowed portions of your image, for example. If you are working on an image that has linear data, then curves can be used to adjust individual channel exposure and essentially be used to perform white balance channel scaling (as long as the curve remains a straight line - i.e., dragging the white point left or down).

Curves and Levels are also good dummy adjustment layers for tonal remapping - if you have highlight data that you want to reinforce, you can add a Curves layer with no adjustment to the points and change the blend mode to Multiply to bring back highlights - add a luminosity mask targeting highlights and you now have more uncompressed highlight data. This is more efficient than duplicating a full pixel layer and changing the blend mode and saves file size. Of course once you use the dummy layer to do whatever it is you want, you can also adjust the curve to tweak the result - for example, add contrast to the reinforced highlights - because it is masked with a luminosity mask, you do not need to worry about the deleterious effect it might have on the midtones and shadows. Because it is an adjustment layer, you can alter the opacity to get the amount of the edit just right. And there are the Blend If sliders to play with too.....

Curves, and specifically a Curves adjustment layer, can do a lot of things, from simple additions of contrast and color to complex remapping of tones and colors with different blend modes, masks and blend ifs. Have fun experimenting!

Kirk

Heya,

I'll make this as simple as it can be regarding curves and why its powerful and meaningful to learn to use:

1. Use the sampler tool on a shadow area and note its value range in RGB in the Info panel (sample size 11 by 11 average; all layers).
2. Sample a few areas and get a range of values and write them down (ie, the complete range of a shadow area from its darkest to brightest (eg, 8 to 16, etc, any values).
3. Make a curve layer. Add two anchor points that envelope that range you measured with the sampler tool (output & input) (eg, 6 to 18 compared to above, or any values).
4. Now make more points around those points to anchor the curve so that adjustments do not effect blacks and whites (they will have limits, around 2 or 3 from each anchor value). Make a bunch to isolate the range.
5. Now gently lift or drop the curve between your measured anchor points and watch the targeted range of shadows get lifted, or insert whatever range you sampled for example without hurting blacks and whites.
6. If adjusting the curve touches your blacks/whites and doesn't just effect the targeted area, add more anchor points outside the measured ranges until the curve doesn't move outside of your targeted area.

Regarding levels:

1. Create a levels layer.
2. Note the data on the left and right, representing black and white points.
3. Stretch the histogram or compress it, by changing where the black point is, and white point is.
4. Anything more advanced than that, and you probably should just target it with a Curve adjustment instead.

++++++++++++

Quick example showing a curve layer used with targeted anchor points based on the values in the info panel to isolate a specific shadow area, and anchor points to prevent the curve adjustment from harming black point or white point or ranges outside of the targeted shadow range. Here I just looked at the deep shadows in the tree line. I raised them slightly, without raising any others, etc. Just to show the idea being suggested here:

MalVeauX wrote in post #18372260
Heya,

I'll make this as simple as it can be regarding curves and why its powerful and meaningful to learn to use:

1. Use the sampler tool on a shadow area and note its value range in RGB in the Info panel (sample size 11 by 11 average; all layers).
2. Sample a few areas and get a range of values and write them down (ie, the complete range of a shadow area from its darkest to brightest (eg, 8 to 16).
3. Make a curve layer. Add two anchor points that envelope that range you measured with the sampler tool (output & input) (eg, 6 to 18 compared to above).
4. Now make more points around those points to anchor the curve so that adjustments do not effect blacks and whites (eg, 4 or 5, and 19 or 20).
5. Now gently lift or drop the curve between your measured anchor points and watch the targeted range of shadows get lifted, or insert whatever range you sampled for example without hurting blacks and whites.
6. If adjusting the curve touches your blacks/whites and doesn't just effect the targeted area, add more anchor points outside the measured ranges until the curve doesn't move outside of your targeted area.

Regarding levels:

1. Create a levels layer.
2. Note the data on the left and right, representing black and white points.
3. Stretch the histogram or compress it, by changing where the black point is, and white point is.
4. Anything more advanced than that, and you probably should just target it with a Curve adjustment instead.

Very best,

I don't understand how you get an RGB value as a single figure as mine are something like 4,5,0 to 26,28,1. Therefore I don't understand how you can set them as an anchor point.

I think he means read the values of each channel: in the info box, while using the sampler, it will show the range of R, G, B seperately.

john crossley wrote in post #18372340
I don't understand how you get an RGB value as a single figure as mine are something like 4,5,0 to 26,28,1. Therefore I don't understand how you can set them as an anchor point.

Sorry, semantics. The values you see next to RGB in the sampler/info box, like 4, 5, 6 or whatever. Set the anchor to something lower (if you can; I don't think you can go lower than 4). Do the same on the other end of a range. Set more anchor points around those so that the curve doesn't change at the black/white side of the curve, and you're effectively isolating a range to then lift or drop.

There are limits to the range of values, something like 2 or so (ie you can't choose 4 and 4, then 5 and 5, more like 7 and 7 or higher, it will prompt you if you're too close with an anchor point and give you a range as a prompt/error report).

Very best,

To place an anchor at a specific RGB value, select the little hand tool thing in the Curves dialog window (the Targeted Adjustment tool) and CMD-SHIFT-click (CTRL-SHIFT-click on PC probably) on the area/tone/pixels in the image that you want to set as the anchor - this will add a control point at the corresponding value of the individual channel on all component channels of the Curve.

Using John Crossley's example point above, if you CMD-SHIFT-clicked on that spot, a control point would be placed at an input of 26 in the Red channel curve, 28 in the Green channel curve and 1 on the Blue channel curve.

If you want to convert the RGB triplet into a single luminosity value, use the rough formula:

0.3*R + 0.59*G + 0.11*B

where R, G, and B are the 8bit values used, for example, in John's info panel above. Using his values (R=26, G=28, B=1), the resulting input point in the composite RGB curve would be:

0.3*(26) + 0.59*(28) + 0.11*(1) = ~24.

If you want the clicked point to be added to only the active curve, like the composite RGB curve, just CMD-click on the area.

kirk

john crossley wrote in post #18372459

[ HOSTED PHOTO DISPLAY FAILED: ATTACH id 859133 has been deleted. ]

So how do you create an anchor point in curves using those three numbers?

You can use the method described below (convert it to luminosity, which is basically what we're doing here, looking at luminosity).

Or, you can just guestimate it, by assuming it's probably close to the 26 & 28 value, like 22~24 as an anchor point, because it will contain those above values for the most part.

All you need is to know a range, like you pointed out, and set anchors that are outside those value by a little bit, then isolate them, and you can target that range.

You can get precise if you convert to luminosity, but you really don't have to for the most part.

I do this for nebulosity lifting and isolating to lift ranges of luminosity to reveal nebulosity. But you can use the same idea to lift shadows in landscape, or to target shadows in a portrait, etc. The point is just the underlying control you have by simply knowing how a curve layer adjustment can work with anchor points for someone wanting to learn more.

Very best,