Learning Landscapes - Please post your BIG picture dramatic landscape images

[Kaye Leggett]

Hi Christina - as you know I am no expert. So maybe set yourself a goal like I did last year - or rather my son did for me. He bought me a beautiful picture frame and told me to fill it..................... So far I have had two pictures in it





So two very different photos of big landscapes.

I like them enough to put them on my wall, that's all that really matters. I haven't asked the family if they like them, they're for me. Go for it.

[Brownbear]

Hi Kaye,

Thank you for sharing. Both are beautiful. The first is something that I could try. I just need to find a field of flowers or buy some and plant them so they bloom in spring.

Hi Christina - as you know I am no expert. So maybe set yourself a goal like I did last year - or rather my son did for me. He bought me a beautiful picture frame and told me to fill it..................... So far I have had two pictures in it





So two very different photos of big landscapes.

I like them enough to put them on my wall, that's all that really matters. I haven't asked the family if they like them, they're for me. Go for it.

[Brownbear]

Hi Mike,

I gave it a try. These I processed in Adobe Photoshop. In ACR I left the WB as shot for the warm image, and auto for the cooler image. Then I adjusted the white and black points, a wee bit, and that is all. (no sharpening)

Then I opened in Adobe Photoshop (note the correct terminology ). Once there I sharpened the entire image .3 radius at 200 percent. Then I applied a curves layer to the mountains, and made a little S curve in the midtones, and then I applied a curves layer to the clouds and made a big S curve (moving the midtone point only)... With the clouds I found it very hard not to make the clouds unnatural looking, ie; hot spots.

Then I cropped and resized to 1200 pixels and applied an unsharp mask of .3 radius at 40 percent.

Is making an S curve to the midtones is the same as applying LCE via an unsharp mask?

Still not the great but I think they are improved. If not, do let me know so I can learn from this. Thank you.



Yes, I was lazy and didn't clone out the weeds in the lower left hand corner. Cleaning up dust spots tired me out.

Christina,

I don't think your image looks at all "blah."

However, I do think you can make it more interesting. For me, most of the sky doesn't have the leading lines that you apparently hoped for (that element is not necessary for my enjoyment). So, try cropping to eliminate about half of it, retaining the part that does have strong leading lines.

More important to me, try applying Local Contrast Enhancement to the forest, increased contrast in the mid tones separately in both forests and also the sky, and a touch more warmth to the forest on the right.

Just a few ideas that you might like.

[Mike Buckley]

a big S curve (moving the midtone point only)

Your comment gives me the impression that you don't understand that making an S-curve results in a shape that at least partly resembles the shape of the letter "S." I think that because it's not physically possible to create an S-curve moving only one point of the curve.

To make an S-curve, whether dramatically or otherwise, you need to drag the starting curve downward about one-fourth of the way in from the left side and upward about one-fourth of the way in from the right side. Doing so changes the diagonal line on the graph to something that resembles the shape of an S.

making an S curve to the midtones

Rather than thinking of "making an S-curve to the midtones," think of applying an S-curve to all of the tones. Doing so increases the contrast of the mid-tones. Notice that the shape of the curve is a bit more vertical in the midtones. Any time the curve is more vertical, there is more contrast.

If you haven't yet figured out what the X and Y axis on the graph of the curve indicate, all of the above will be meaningless. I'm getting the impression that you haven't come to an understanding of that yet. Otherwise, I don't think you would be asking the questions in the way you are asking them.

is making an S curve to the midtones the same as applying LCE via an unsharp mask?

No, not at all. Look up the CiC tutorial about LCE and you'll probably understand why that's the case.

[Brownbear]

Hi Mike,

Thank you for advising. Perhaps I'm still a little confused and I'm also having a hard time seeing the S in the curve. For some reason I was under the impression that increasing the midtones (moving the middle part of the curve up, instead of darkening the midtones = lowering the center portion of the curve down?) results in a version of an S curve and this increases contrast, unless it is a reverse S curve.

In the tutorial it is fairly easy to see the shapes but not so easy to see when one is trying to apply a unique S curve, and I thought I was applying an S curve because I increased the midtones (lightened them).

I did check the tutorial on local contrast enhancement and I have used this technique before for sharpening, but I thought that there was more than one way to do these things, and since I'm increasing the contrast using the curves tool is it not the same as sharpening (because it has an effect on the tones in the image) indirectly by increasing the contrast? (Even though there is no radius tool) If not then I will try this edit again tomorrow.

I would appreciate it very much if you could look at my S curves applied and steer me in the right direction.

Here is the curve to the mountain



Here is the curve to the clouds




Thank you.

Your comment gives me the impression that you don't understand that making an S-curve results in a shape that at least partly resembles the shape of the letter "S." I think that because it's not physically possible to create an S-curve moving only one point of the curve.

To make an S-curve, whether dramatically or otherwise, you need to drag the starting curve downward about one-fourth of the way in from the left side and upward about one-fourth of the way in from the right side. Doing so changes the diagonal line on the graph to something that resembles the shape of an S.



Rather than thinking of "making an S-curve to the midtones," think of applying an S-curve to all of the tones. Doing so increases the contrast of the mid-tones. Notice that the shape of the curve is a bit more vertical in the midtones. Any time the curve is more vertical, there is more contrast.

If you haven't yet figured out what the X and Y axis on the graph of the curve indicate, all of the above will be meaningless. I'm getting the impression that you haven't come to an understanding of that yet. Otherwise, I don't think you would be asking the questions in the way you are asking them.



No, not at all. Look up the CiC tutorial about LCE and you'll probably understand why that's the case.

[Mike Buckley]

This CiC tutorial displays two typical S-curves in the discussion under the heading, "Motivation: Dynamic Range & Film Curves." To see the first of those two S-curves, hover your mouse over the first graph. The second S-curve is the one displayed in blue.

The very first curve displayed in that tutorial under a different heading is also a classic S-curve.

Notice that those three curves are roughly in the shape of an S. Your first curve isn't at all in the shape of an S. Your second curve is more in the shape of a reverse S-curve.

I was under the impression that increasing the midtones (moving the middle part of the curve up, instead of darkening the midtones = lowering the center portion of the curve down?) results in a version of an S curve and this increases contrast

Nope, not at all.

To repeat: anywhere the shape of the curve is more vertical than the starting shape of the curve, that area in the tonal range will display more contrast. The middle section of an S-curve is the part that is most vertical. Thus, the tonal range in that area, which happens to be the mid-tones, will display increased contrast.

Do you feel comfortable that you fully understand the data represented in the X and Y axis? If not, we really need to attend to that. Otherwise, all discussions about the Curve tool are meaningless.

[ajohnw]

I haven't looked at the tutorial but inevitably they show an S curve. The curve is dragged down lower at the low dark end and pushed up in the upper bright end. These 2 points make the curve look like a tilted S. The net effect of this is to make the slope of the curve steeper between the 2 points and that will increase contrast in the range of tones between them. The slope is reduced at each end though which will reduce contrast in those regions. The same effect can be achieved by setting a black point. That effectively move the bottom left end of the curve across and again increases the slope of the curve. It also forces any tone that are lower than the point to black.

Your curves are fine - just bear in mind the slope effect. You can use any shape of curve you like. Some will cause all sorts of problems but that is the best way to learn.

One approach is to realise that some contrast loss some where is inevitable and restore it with tone mapping if you have it. In other words you might use a tone curve from raw which leaves things a bit flat and then bring it up with tone mapping. Tone mapping increases contrast and how it's applied varies across packages. Some brighten as well so the image may need darkening. One I use applies a contrast boost on the basis of how contrasty the image is locally allowing low contrast areas to be boosted more than higher contrast areas.

The contrast slider can be used as well. The problem with this sort of thing is that all of the PP steps inter reacts and it takes a while to pick it up.

On the 2 screen shots you have posted you may be able to select the mountains etc some how and just apply a tone curve to those and then invert the selection and apply an entirely different one to the rest. The only other way to do this is to use HDR techniques. Generate what are effectively 2 exposures and then blend to 2 together. In shots like the one that you have posted that would usually mean manipulating one exposure where all but the clouds are near black. The part of the curve that covers the clouds can then be very steep. The other exposure in simple terms would show the mountains and trees etc as you want more or less neglecting the clouds. The 2 are then blended together. The main problem is that the tones in the 2 exposures inter react so the black mountains in the one will darken the lighter ones in the other. All good fun but it can be done.

John
-

[Brownbear]

Hi Mike,

Yes, I can see that from the tutorials, but I will read them a couple more times and hopefully this time around it will sink in more.

I was applying what I thought was an S curve to just the middle portion of the curve, and I can see a small S in the first adjustment, and maybe not in the 2nd adjustment.

Thank you for clarifying that. (LCE vs curves)

I think I have a basic understanding. The X axis represents all the tones in an image from black to white, as does the y axis, with the basic curves line down the center untouched being the middle grey value. The y axis shows the change in the tonal values... ie; the point on the line at the bottom lifted up, lightens the shadows (y) and a point on the y axis lifted up in the highlight region represents the lightened shadows, and if I drag a middle midpoint down it represents (y) a darker midtone.

So if I raise a point near the highlights and lower a point near the shadows I am increasing the contrast. And I thought that if I did the same thing to the middle of the curve it would also increase the contrast (s curve in the mid tones)

?

This CiC tutorial displays two typical S-curves in the discussion under the heading, "Motivation: Dynamic Range & Film Curves." To see the first of those two S-curves, hover your mouse over the first graph. The second S-curve is the one displayed in blue.

The very first curve displayed in that tutorial under a different heading is also a classic S-curve.

Notice that those three curves are roughly in the shape of an S. Your first curve isn't at all in the shape of an S. Your second curve is more in the shape of a reverse S-curve.



Nope, not at all.

To repeat: anywhere the shape of the curve is more vertical than the starting shape of the curve, that area in the tonal range will display more contrast. The middle section of an S-curve is the part that is most vertical. Thus, the tonal range in that area, which happens to be the mid-tones, will display increased contrast.

Do you feel comfortable that you fully understand the data represented in the X and Y axis? If not, we really need to attend to that. Otherwise, all discussions about the Curve tool are meaningless.

[Brownbear]

I haven't looked at the tutorial but inevitably they show an S curve. The curve is dragged down lower at the low dark end and pushed up in the upper bright end. These 2 points make the curve look like a tilted S. The net effect of this is to make the slope of the curve steeper between the 2 points and that will increase contrast in the range of tones between them. The slope is reduced at each end though which will reduce contrast in those regions. The same effect can be achieved by setting a black point. That effectively move the bottom left end of the curve across and again increases the slope of the curve. It also forces any tone that are lower than the point to black.
John,

Thank you for a very helpful explanation of curves. I just played with the curve tool on an image in Lightroom and I can see where the slope of the line is steeper in the middle area and less steep in the highlights and shadows section. And moving the white and black point makes it easier to see the steepness of the slope, hence contrast. Which is reinforcing what Mike is trying to help me understand.

Pointing out the loss of contrast on other points of the curve helps reinforce the material in the tutorials.

Right now just learning the curves tool is more than enough for me. I seem to be a little slow on learning post processing, so one step at a time will do me just fine.

Thank you. Truly appreciated.


Your curves are fine - just bear in mind the slope effect. You can use any shape of curve you like. Some will cause all sorts of problems but that is the best way to learn.

One approach is to realise that some contrast loss some where is inevitable and restore it with tone mapping if you have it. In other words you might use a tone curve from raw which leaves things a bit flat and then bring it up with tone mapping. Tone mapping increases contrast and how it's applied varies across packages. Some brighten as well so the image may need darkening. One I use applies a contrast boost on the basis of how contrasty the image is locally allowing low contrast areas to be boosted more than higher contrast areas.

The contrast slider can be used as well. The problem with this sort of thing is that all of the PP steps inter reacts and it takes a while to pick it up.

On the 2 screen shots you have posted you may be able to select the mountains etc some how and just apply a tone curve to those and then invert the selection and apply an entirely different one to the rest. The only other way to do this is to use HDR techniques. Generate what are effectively 2 exposures and then blend to 2 together. In shots like the one that you have posted that would usually mean manipulating one exposure where all but the clouds are near black. The part of the curve that covers the clouds can then be very steep. The other exposure in simple terms would show the mountains and trees etc as you want more or less neglecting the clouds. The 2 are then blended together. The main problem is that the tones in the 2 exposures inter react so the black mountains in the one will darken the lighter ones in the other. All good fun but it can be done.

John
-

[Mike Buckley]

The X axis represents all the tones in an image from black to white, as does the y axis

Correct. However, there is more to it than that, as explained below.

with the basic curves line down the center untouched being the middle grey value.

No.

That diagonal line is simply a graphical interface that allows you to remap (that word, "remap," is vitally important) the tones from what they are to what you want them to be.

Imagine that you want to remap all pixels with a tonal value of 100 to a tonal value of 110. Ignore for the moment that it is not possible to remap just one pixel without affecting the pixels of similar tonal values. If you could do that, you would hover your mouse over the diagonal line until it was at the value of 100 on the X axis. You would then drag that one tiny part of the diagonal line (nothing else) upward until it reaches the value of 110 on the Y axis.

Now back to reality: You can't just drag the value of 100 to a value of 110; the other values to the left and right of 100 are also dragged upward though not to the same degree.

So if I raise a point near the highlights and lower a point near the shadows I am increasing the contrast.

Correct. But you didn't explain the tonal range that incurs that increase in contrast, which is really important to understand. You also didn't explain the tonal range that incurs a decrease in contrast, which is equally important to understand. Think about that and get back to us.

As you think about it, consider that the CiC tutorial explains increased contrast in the context of "stretched" tones and decreased contrast in the context of "compressed" tones, which, frankly, I have never understood. (I'm not suggesting that the information is wrong. Instead, I don't understand the concept of stretched and compressed tones. Hopefully you or someone will be able to explain it to me.)

All of that is important because it's physically impossible to increase contrast in one area of the curve without decreasing contrast in another area. We need to consciously make those decisions when we alter the curve to be able to achieve the look that we desire.

I thought that if I did the same thing [applied an S-curve] to the middle of the curve it would also increase the contrast (s curve in the mid tones)

Remember that applying an S-curve always increases the contrast in one area of the tones and decreases the contrast in other areas. Once you answer the above questions about which parts of the tone curve incur increased contrast and which parts incur decreased contrast, you'll understand why applying an S-curve to just one small part of the overall curve does not increase the contrast throughout that part of the curve.

Hang in there! You'll get it. And when you do, you'll need to replace all of the light bulbs in your house because all of them will be glowing so bright they'll explode.

[Brownbear]

Hi Mike,


The x axis represents the tones black, grey to white (zone system) before I make any changes to the curve = input tones

The y axis represents the tones black, grey to white (zone system) after I make any changes = output tones


So when I lift sections of the curves tool up I am lightening the tones in that area (Shadows, midtones or highlights) and when I pull sections of the curves tool down I am darkening the tones in that area... And anywhere I increase the slope of the curve = increased contrast in that area, and less contrast in other areas. In the case of the S curve this would be increased contrast (lightening the midtones greys) at the expense of contrast in the shadows and highlights area.

Therefore when I use the curves too I'm brightening or darkening select tones in my chosen area?

Thank you.

Correct. However, there is more to it than that, as explained below.



No.

That diagonal line is simply a graphical interface that allows you to remap (that word, "remap," is vitally important) the tones from what they are to what you want them to be.

Imagine that you want to remap all pixels with a tonal value of 100 to a tonal value of 110. Ignore for the moment that it is not possible to remap just one pixel without affecting the pixels of similar tonal values. If you could do that, you would hover your mouse over the diagonal line until it was at the value of 100 on the X axis. You would then drag that one tiny part of the diagonal line (nothing else) upward until it reaches the value of 110 on the Y axis.

Now back to reality: You can't just drag the value of 100 to a value of 110; the other values to the left and right of 100 are also dragged upward though not to the same degree.



Correct. But you didn't explain the tonal range that incurs that increase in contrast, which is really important to understand. You also didn't explain the tonal range that incurs a decrease in contrast, which is equally important to understand. Think about that and get back to us.

As you think about it, consider that the CiC tutorial explains increased contrast in the context of "stretched" tones and decreased contrast in the context of "compressed" tones, which, frankly, I have never understood. (I'm not suggesting that the information is wrong. Instead, I don't understand the concept of stretched and compressed tones. Hopefully you or someone will be able to explain it to me.)

All of that is important because it's physically impossible to increase contrast in one area of the curve without decreasing contrast in another area. We need to consciously make those decisions when we alter the curve to be able to achieve the look that we desire.



Remember that applying an S-curve always increases the contrast in one area of the tones and decreases the contrast in other areas. Once you answer the above questions about which parts of the tone curve incur increased contrast and which parts incur decreased contrast, you'll understand why applying an S-curve to just one small part of the overall curve does not increase the contrast throughout that part of the curve.

Hang in there! You'll get it. And when you do, you'll need to replace all of the light bulbs in your house because all of them will be glowing so bright they'll explode.

[Mike Buckley]

Christina sent me the same information provided in her most recent post as a private message, so my response shown below is a copy of the private message that I sent to her.

==================================

Almost perfect, Christina! Very good!

A minor mistake at least in the way that I interpret your statement is that the tone values represented in each axis are not the zone system. The zone system by Ansel Adams began as an analog system of 11 tones. Each axis represents 256 tones. Even so, if you were trying to imply that all of the 11 tones represented in the zone system are included in the 256 tones represented on each axis, you are correct.

In the case of the S curve this would be increased contrast (lightening the midtones greys) at the expense of contrast in the shadows and highlights area.

Think of an S-curve in which the shape on the left half of the curve is an exact mirror image of the shape on the right half of the curve. In that situation, the very middle part of the mid-tones around the value of 128 remain relatively unaltered. The brighter part of the mid-tones are slightly lightened but we wouldn't say that the mid-tones in general are lightened. That's because the darker part of the mid-tones are actually slightly darkened. (You can easily see that now that you know how to pay detailed attention to the input values represented on the X axis and output values represented on the Y axis.)

The details in your statement about increased and decreased contrast are absolutely correct.

Now, to put your new-found knowledge to practical use: Let's assume that you want to brighten your image overall without using a "brighten" slider. How would you do that using the Curve tool?

Once you have done that, keeping in mind that any alteration of the curve increases contrast in at least one area and decreases contrast in at least another area, which area(s) incur increased contrast? Which areas incur decreased contrast?

You are almost there to completely understanding how to use the Curve tool in real-world situations. Once you're there, a whole new world of control will be open to you that far exceeds any control any slider that affects brightness or contrast can ever provide.

[ajohnw]

Mike has explained this area a little clearer then I did. There is one other way of looking at curves which might help. Just assume black and white but the curve changes all tones at the same level irrespective of colour. It's easier to explain if numbers are used.

Some explanations state that the graph basically shows input on one axis and output on the other. Zero tone rgb=0,0,0 is at bottom left. Max tone 255 is top right. The default line runs from corner to corner = no change. Tone can be regarded as brightness so say an anchor point is placed at 127,127. If that is raised the image will be brightened most at that point trailing off to no change at 0 and 255. Contrast is the rate at which brightness changes so as the slope of the line has increased in the 0 to 127 region contrast has also increased. From 127 to 255 the slope has decreased so contrast is reduced even though the range has been brightened. If the 127 position is moved down darkening the image the opposite happens. Slope is reduced from 0 to 127 and increased from 127 to 255.

The catch is that nothing can be darker than 0 or brighter than 255. This why contrast increases in one area must be at the expense of contrast in another area. We are stuck with the brightness / tone range the numbers represent. Actually when working with the usual RGB's the range is 0 to 255 and only deep colour gamuts with more than 8 bits per channel can offer more. The knock on effect of this is that changes can only be carried out so far because at some point local retouching can't cure the problems the curve change might introduce..

Using curves has to relate to the shot that's been taken. It's easier to discuss the cloud shots in the other post. The aim here is the clouds so decreasing the slope of the curve in the tone/brightness range these use would be a bad idea. It may need increasing until the clouds look as required. One simple way to do that is too look at the visual information in the shot. There is very little in the boat so that tone range can be dragged down reducing it further - find the point on the curve that changes that tone level and drag it down or set it as the black or near black point by shifting the 0,0 point across horizontally. In some situations the same thing might be done and the areas that suffers retouched a little locally. In that case removing all of the tonal variation via a black point wouldn't be a good idea as pure black can only be changed to a flat shade of grey. This is where the bottom dark end of typical S curves come in. They leave some tonal variation that is ideally perfect when applied from raw but if not leaves some scope for local retouching. This sort of thing is often needed on subjects such as the landscape you posted in this thread as the dynamic/brightness range in the scene is to large for the camera to capture well.

Another common problem is that some area of a shot is too dark to show detail well or an area just needs brightening a little. Often that can be fixed by simply shifting the right point on the curve up.If taken too far the bad effects will be obvious. Contrast changes might restore that aspect but only by so much. Too much brightening can also cause colour artefacts to appear. Local retouching can be a better option.

The shot in this post is more difficult. The best way would be to do the curve work from raw. Usually it's best to ensure that the highlights are correct at that stage because they can be rather difficult to fix later. The shadowy areas can be bought up later slowly locally with a burning brush. This sort of thing might mean leaving the black point at zero and dragging a point down say 1/3 ( or some where else) of the way across to increase contrast above it. Local retouching can then be be used to bring up detail in the darker areas where needed. Or an S curve which might sacrifice very little highlight detail. There is also an approach where the shot from raw is deliberately left a little dark so that highlights can be brightened or the contrast changed later without clipping. The raw's exposure compensation adjustment is probably the best way to try that.

The reason for trying to get highlights right is that any further processing might clip them as they can't go higher than 255. The dark end on the other hand can always be brightened / changed in a number of ways locally with touch up brushes as there is headroom available. Clouds or other bright areas might reach 255 or get very close to it and can't be changed further only darkened.

The histogram is useful for getting an idea of the level of visual information at various levels and if there is space for adjustment. It goes dark to bright tones in the same way that the curves scales do. Height corresponds to the level of visual detail. The rgb ones are less important than the luminance one for this use - but the image itself is the final metric.

On thing I noticed on the shots posted in this thread is that I couldn't sensibly brighten the dark areas in the tree line. This suggests that the tone levels there are clipped to black ( 0,0,0). This might be down to a default setting in your raw converter. If this area had been covered by the bottom part of an S curve it's usually possible to bring up some shadowy detail / colour via local retouching even by just using a burning / brightening brush. Packages may have brushes for contrast, saturation or even tone mapping as well as dodging (darkening) and burning brushes. Also other features for removing unwanted items in the shot. The black clipping might also be down to the camera / exposure used. It varies but as far as clouds are concerned I find that it's best to expose so that the jpg preview shows very little or no clipping. Many cameras have the facility to make clipped areas blink in the preview. If they do set the minimum exposure compensation to remove them and take the shot again. It's also possible to use the histogram display in the viewfinder and note that the highlights do trail off to zero and don't end abruptly. Not so good as taking a shot to see as it may not be metering the entire cloud area. Some people play with jpg contrast settinhgs to try and gauge actual raw exposures as there is usually or used to be some head room in the raw file over what the jpg could capture. Many cameras now apply an S curve when they convert to jpg. Canon never did but have joined the club of late. Some cameras offer several different curves for jpg conversion. One of these may capture every scrap of highlight capture available in raw or make a serious dent in it. It's best to take some shots and see what happens. Some small areas of over exposed clouds may be recoverable from raw It all depends on the curve the camera uses to produce it's jpg's. One thing is for sure. Clipped highlights in raw files can not be recovered full stop. Leaving a bit of lee way isn't a bad thing to do. If the other end finishes up too dark to allow the intended results to be obtained the only cure is 2 or more shot HDR techniques. The results of that needn't look like typical HDR shots.

Summary - scenes like this one need curves adjustments and local retouching. Hope there are no typos but this one is a bit long and I went out for several hours since starting and finishing it.

John
-

[Mike Buckley]

Nice post, John.

You mentioned that it often becomes necessary to locally make changes after globally adjusting the curve. One way to do that is to select an area and to use the Curve tool to make the desired changes.

You also mentioned of the need to keep an eye on the histogram when adjusting the curve. That's worth repeating for emphasis. I think it's important to constantly keep an eye on the histogram regardless of the alteration being made to a photograph.

[ajohnw]

I mostly feel that brushes are a better option than selection as selection can take some time depending on what type and then there are blend effects etc. The main thing with brushes is to set them up so that the effect builds up slowly.

I use a Linux package called Fotoxx a lot. Any change can be applied via a brush or a selection. It's just changed it's presentation of curves. It was along with R G B and saturation a straight horizontal line above a grey scale. Also an array of button to lift lower or tilts the line either way and the ability to add as many points as needed. Curious thing is that I have been using that with no thoughts about contrast changes at all for some time now. They can be seen when they occur and corrected in other ways, usually a brush. The author has now changed to a convention curve arrangement as the old one implied white could go whiter than white etc. I'm trying to get him to put it back as it was again. Even though I can cope with curves from hours with the gimp I prefer the other way. One nice thing is that once the tone range is right the whole thing can be moved up and down with the move up or down buttons to finally tweak the brightness. To be honest I feel that this is a far better arrangement than any other package I am aware of. They all use the curve approach. As things stand with the new set up I have to adjust brightness by moving the entire curve up which doesn't make any sense at all for obvious reasons. It can't go up or down due to the end points. Rawtherapee gets round this by having a "lightness slider". What else could they call it.

It might be worth adding some camera jpg curves. Curves for most packages can be found on the web as well. Usually with a description of what they do to usually raw files. This is the EM-5 jpg curve set. 0 to 255 against stops this time as the colour channels are 12 bit each in raw not 8 as in the jpg. There are also another set of curves that work along side these and provide minor tweaks.



There is then yet another curve to adjust these via the viewfinder to change the appearance in it before the shot has even been taken.. As you can see assuming that there really is more highlight in a raw file can be a bit dubious. Canon's can tend to be a bit of a different kettle of fish. Not so much Nikons. All I have used on the EM-5 so far is normal.

People can find out what curves their camera has by searching dpreview "make model" and looking at dynamic range in the full reviews. At least 9 stops is fairly normal but there are options for increasingly more these days.

How these relate to out of camera eg Adobe raw conversion? I have been trying to find out but no one seems to know. Adobe mention tone mapping into 8 bit space which is what these camera curves are doing but they don't give much of an explanation - of no concern to the user.

John
-

[Mike Buckley]

I mostly feel that brushes are a better option than selection

In my mind, brushes are a type of selection.

[Brownbear]

Firstly, thank you to Mike and John for helping me gain a better understanding of the curves tool.

Mike...

Thank you for taking the time to provide a visual for the S curve and explaining the relationship of the Zone system.

I feel a little bit overloaded with information but I will try to answer your question on brightening an image.

To brighten the image I could choose to lift up a point anywhere on the curve. If I lift up a point on the curve in the midtone area it decreases the slope in the shadows and highlights area resulting in less contrast in these areas, and brightening the shadows or the highlights results in less contrast in the midtones.

So as it relates to the image that I posted in this thread, is it that I should brighten the trees using the curves tool or increase the contrast in the tree area just using local contrast enhancement. And if I were to brighten the trees just how does one know which point on the curve to lift up... Do I just pick a green that I wish to make brighter and guess which point that it on the curve? And would it not make sense to go into the green tone area and not touch the reds and the blues? And when you look at clouds how do you decide if they need more contrast? I guess I am asking how you can just look at my image and be able to advise me specifically what area needs to be worked on and how.

Where are you finding these input and output values? Can I read about them somewhere? How does the curves tool relate to the histogram?

John...

Thank you for a very thorough explanation and relating them to some of the landscape images I have recently taken, very helpful. I see you have a numbers guide for the tones.

I am going to have to read your reply a few times because it is loaded with lots of great information that I will need a little time to think it through.

With respect to the image in this thread that I am going to try and post process to improve (once I am clear on what I am doing).. I have two more shots of it at different exposures. In this one I exposed for the clouds (nothing clipped except by my post processing) and I have another exposed for the trees, and a middle one. I was experimenting with bracketing. (but I'm not ready to try putting photos together). The land mass is not clipped but very dark indeed so when I try another edit I will pay careful attention to see how much I can lighten it without degrading the image.

Thank you to both.

Aside... This thread has evolved to include a lesson in understanding curves for me (and hopefully others, too)... No matter keep posting your landscapes.

[Mike Buckley]

To brighten the image I could choose to lift up a point anywhere on the curve. If I lift up a point on the curve in the midtone area it decreases the slope in the shadows and highlights area resulting in less contrast in these areas, and brightening the shadows or the highlights results in less contrast in the midtones.

Mostly good response but let's clarify a couple things.

The challenge was not to brighten the image but was to brighten the image overall. The ideal way to do that is basically as you suggested but more specifically to select the input value of 127, which is exactly in the middle of the tone values, and drag it upward.

To explain where the revised curve will incur increased and decreased contrast, I'll simply copy and paste from John's post:

"Tone can be regarded as brightness so say an anchor point is placed at 127,127 [those are the input and output values before any change has been made]. If that is raised the image will be brightened most at that point trailing off to no change at 0 and 255. Contrast is the rate at which brightness changes so as the slope of the line has increased in the 0 to 127 region contrast has also increased. From 127 to 255 the slope has decreased so contrast is reduced even though the range has been brightened. If the 127 position is moved down darkening the image the opposite happens. Slope is reduced from 0 to 127 and increased from 127 to 255."

So as it relates to the image that I posted in this thread, is it that I should brighten the trees using the curves tool or increase the contrast in the tree area just using local contrast enhancement.

That depends on what you're trying to accomplish.

And if I were to brighten the trees just how does one know which point on the curve to lift up...

Most software will allow you to hover your mouse over the trees and display the corresponding tonal value as a black dot on the Curve tool's diagonal line.

And would it not make sense to go into the green tone area and not touch the reds and the blues?

You could adjust the tone curve of just the red channel, or the green channel, or just the blue channel, if your software allows you to do that. All of that stuff is pretty advanced and is often used to adjust the color balance if you apply the revised curve to the entire image.

And when you look at clouds how do you decide if they need more contrast? I guess I am asking how you can just look at my image and be able to advise me specifically what area needs to be worked on and how.

That's just judgement; everyone will have their own opinion and nobody's opinion is right or wrong. The judgement comes from experience of learning how you want your images to look. You've been doing this for relatively little time and my guess, based on all of your other advancements, is that a year from now you'll be past that part of your learning curve.

Where are you finding these input and output values?

When using my software, I can hover my mouse anywhere in the Curve tool's graph. The software automatically displays the corresponding input and output values in fields below the graph.

How does the curves tool relate to the histogram?

Great question! The X axis of the histogram displays the same data as the X axis of the Curve tool. The Y axis of the histogram displays the relative number of pixels of a particular tone value in the image.

So, consider this example: Approximate the middle of the X axis as being tone values near 127. If that part of the histogram extends higher up the Y axis than any of the other tonal values, you know that there are far more mid-tones in your image than dark or bright tones.

When you adjust the tone curve, the histogram will always change at least somewhat. That's because adjusting the tone curve remaps (remember that important word?) some if not most of the tonal values. Doing so changes the distribution of tones, so those changes are reflected in a different histogram.

Indeed, some Curve tools display the histogram within the Curve tool. When that's the case, the Y axis means one thing when viewing the curve and another thing when viewing the histogram.

[Brownbear]

Hi Mike,

Thank you for taking the time to answer all of my questions and add clarification where needed. This will be a printer that I need a day or two to digest, and then I will try post processing that image again just for the exercise, hoping it will stand me in good stead for future photos and post processing. Lightroom displays a percentage value for the numbers so I will explore the hovering in Adobe Elements or Photoshop.

Thanks a million, Mike! If there were a BIG hug/thanks a million button I would insert it (several) and send it your way. I expect you need a well-deserved break from me... Go forth and enjoy!

Mostly good response but let's clarify a couple things.

The challenge was not to brighten the image but was to brighten the image overall. The ideal way to do that is basically as you suggested but more specifically to select the input value of 127, which is exactly in the middle of the tone values, and drag it upward.

To explain where the revised curve will incur increased and decreased contrast, I'll simply copy and paste from John's post:

"Tone can be regarded as brightness so say an anchor point is placed at 127,127 [those are the input and output values before any change has been made]. If that is raised the image will be brightened most at that point trailing off to no change at 0 and 255. Contrast is the rate at which brightness changes so as the slope of the line has increased in the 0 to 127 region contrast has also increased. From 127 to 255 the slope has decreased so contrast is reduced even though the range has been brightened. If the 127 position is moved down darkening the image the opposite happens. Slope is reduced from 0 to 127 and increased from 127 to 255."



That depends on what you're trying to accomplish.



Most software will allow you to hover your mouse over the trees and display the corresponding tonal value as a black dot on the Curve tool's diagonal line.



You could adjust the tone curve of just the red channel, or the green channel, or just the blue channel, if your software allows you to do that. All of that stuff is pretty advanced and is often used to adjust the color balance if you apply the revised curve to the entire image.



That's just judgement; everyone will have their own opinion and nobody's opinion is right or wrong. The judgement comes from experience of learning how you want your images to look. You've been doing this for relatively little time and my guess, based on all of your other advancements, is that a year from now you'll be past that part of your learning curve.



When using my software, I can hover my mouse anywhere in the Curve tool's graph. The software automatically displays the corresponding input and output values in fields below the graph.



Great question! The X axis of the histogram displays the same data as the X axis of the Curve tool. The Y axis of the histogram displays the relative number of pixels of a particular tone value in the image.

So, consider this example: Approximate the middle of the X axis as being tone values near 127. If that part of the histogram extends higher up the Y axis than any of the other tonal values, you know that there are far more mid-tones in your image than dark or bright tones.

When you adjust the tone curve, the histogram will always change at least somewhat. That's because adjusting the tone curve remaps (remember that important word?) some if not most of the tonal values. Doing so changes the distribution of tones, so those changes are reflected in a different histogram.

Indeed, some Curve tools display the histogram within the Curve tool. When that's the case, the Y axis means one thing when viewing the curve and another thing when viewing the histogram.

[Brownbear]

Mike has explained this area a little clearer then I did. There is one other way of looking at curves which might help. Just assume black and white but the curve changes all tones at the same level irrespective of colour. It's easier to explain if numbers are used.
Hi John,
Just to say that if their were a BIG thanks a million button I would send you one, too. I've noticed that you have jumped in several times to help me along with my technically deficient brain, and it is truly appreciated. Just so you know.




Some explanations state that the graph basically shows input on one axis and output on the other. Zero tone rgb=0,0,0 is at bottom left. Max tone 255 is top right. The default line runs from corner to corner = no change. Tone can be regarded as brightness so say an anchor point is placed at 127,127. If that is raised the image will be brightened most at that point trailing off to no change at 0 and 255. Contrast is the rate at which brightness changes so as the slope of the line has increased in the 0 to 127 region contrast has also increased. From 127 to 255 the slope has decreased so contrast is reduced even though the range has been brightened. If the 127 position is moved down darkening the image the opposite happens. Slope is reduced from 0 to 127 and increased from 127 to 255.

The catch is that nothing can be darker than 0 or brighter than 255. This why contrast increases in one area must be at the expense of contrast in another area. We are stuck with the brightness / tone range the numbers represent. Actually when working with the usual RGB's the range is 0 to 255 and only deep colour gamuts with more than 8 bits per channel can offer more. The knock on effect of this is that changes can only be carried out so far because at some point local retouching can't cure the problems the curve change might introduce..

Using curves has to relate to the shot that's been taken. It's easier to discuss the cloud shots in the other post. The aim here is the clouds so decreasing the slope of the curve in the tone/brightness range these use would be a bad idea. It may need increasing until the clouds look as required. One simple way to do that is too look at the visual information in the shot. There is very little in the boat so that tone range can be dragged down reducing it further - find the point on the curve that changes that tone level and drag it down or set it as the black or near black point by shifting the 0,0 point across horizontally. In some situations the same thing might be done and the areas that suffers retouched a little locally. In that case removing all of the tonal variation via a black point wouldn't be a good idea as pure black can only be changed to a flat shade of grey. This is where the bottom dark end of typical S curves come in. They leave some tonal variation that is ideally perfect when applied from raw but if not leaves some scope for local retouching. This sort of thing is often needed on subjects such as the landscape you posted in this thread as the dynamic/brightness range in the scene is to large for the camera to capture well.

Another common problem is that some area of a shot is too dark to show detail well or an area just needs brightening a little. Often that can be fixed by simply shifting the right point on the curve up.If taken too far the bad effects will be obvious. Contrast changes might restore that aspect but only by so much. Too much brightening can also cause colour artefacts to appear. Local retouching can be a better option.

The shot in this post is more difficult. The best way would be to do the curve work from raw. Usually it's best to ensure that the highlights are correct at that stage because they can be rather difficult to fix later. The shadowy areas can be bought up later slowly locally with a burning brush. This sort of thing might mean leaving the black point at zero and dragging a point down say 1/3 ( or some where else) of the way across to increase contrast above it. Local retouching can then be be used to bring up detail in the darker areas where needed. Or an S curve which might sacrifice very little highlight detail. There is also an approach where the shot from raw is deliberately left a little dark so that highlights can be brightened or the contrast changed later without clipping. The raw's exposure compensation adjustment is probably the best way to try that.

The reason for trying to get highlights right is that any further processing might clip them as they can't go higher than 255. The dark end on the other hand can always be brightened / changed in a number of ways locally with touch up brushes as there is headroom available. Clouds or other bright areas might reach 255 or get very close to it and can't be changed further only darkened.

The histogram is useful for getting an idea of the level of visual information at various levels and if there is space for adjustment. It goes dark to bright tones in the same way that the curves scales do. Height corresponds to the level of visual detail. The rgb ones are less important than the luminance one for this use - but the image itself is the final metric.

On thing I noticed on the shots posted in this thread is that I couldn't sensibly brighten the dark areas in the tree line. This suggests that the tone levels there are clipped to black ( 0,0,0). This might be down to a default setting in your raw converter. If this area had been covered by the bottom part of an S curve it's usually possible to bring up some shadowy detail / colour via local retouching even by just using a burning / brightening brush. Packages may have brushes for contrast, saturation or even tone mapping as well as dodging (darkening) and burning brushes. Also other features for removing unwanted items in the shot. The black clipping might also be down to the camera / exposure used. It varies but as far as clouds are concerned I find that it's best to expose so that the jpg preview shows very little or no clipping. Many cameras have the facility to make clipped areas blink in the preview. If they do set the minimum exposure compensation to remove them and take the shot again. It's also possible to use the histogram display in the viewfinder and note that the highlights do trail off to zero and don't end abruptly. Not so good as taking a shot to see as it may not be metering the entire cloud area. Some people play with jpg contrast settinhgs to try and gauge actual raw exposures as there is usually or used to be some head room in the raw file over what the jpg could capture. Many cameras now apply an S curve when they convert to jpg. Canon never did but have joined the club of late. Some cameras offer several different curves for jpg conversion. One of these may capture every scrap of highlight capture available in raw or make a serious dent in it. It's best to take some shots and see what happens. Some small areas of over exposed clouds may be recoverable from raw It all depends on the curve the camera uses to produce it's jpg's. One thing is for sure. Clipped highlights in raw files can not be recovered full stop. Leaving a bit of lee way isn't a bad thing to do. If the other end finishes up too dark to allow the intended results to be obtained the only cure is 2 or more shot HDR techniques. The results of that needn't look like typical HDR shots.

Summary - scenes like this one need curves adjustments and local retouching. Hope there are no typos but this one is a bit long and I went out for several hours since starting and finishing it.

John
-