Shooting quilts with flash indoors

[Mike Buckley]

Thanks for the follow-up, Dan.

We do indeed agree for the most part once we get some semantics out of the way. When you were referring to the extremes, I mistakenly thought you were referring to two tones -- the darkest tone and the brightest tone. I now realize that you're referring to the two ranges of extremes rather than only the extremes.

I'll take one of your comments even farther than you did...

That is the reason for the S-curve--it increases tonal range in the mid range, often at the cost of tonal range elsewhere (depending on what the total range in the image is, and how the midrange is expanded).

If there are three ranges of luminosity in the image -- shadows, mid tones and highlights -- it's actually impossible to apply a traditional S-curve to increase the mid-tone contrast without decreasing the contrast in the shadows and highlights.

However, an important point needs clarifying...

The tonal range has been stretched out considerably. This was done with the LR contrast slider, but I would have obtained a very similar histogram by using the blacks and white sliders, which among other things change the black and white points. I would expect the image to show more contrast. And indeed it does

The two histograms you displayed prove that the Lightroom contrast slider moves the black and white points. As you know, doing so inherently adjusts all other parts of the tone curve; it's impossible to adjust the black and white points without adjusting the rest of the tone curve. I suspect that the contrast slider makes additional adjustments to the mid-tone area of the tone curve as well, much as what would happen if a traditional S-curve had been applied. People less informed than you use such sliders not having any idea about what is really happening to the tone curve.

[Mike Buckley]

Thanks for the follow-up, Dan.

We do indeed agree for the most part once we get some semantics out of the way. When you were referring to the extremes, I mistakenly thought you were referring to two tones -- the darkest tone and the brightest tone. I now realize that you're referring to a range of extremes.

I'll take one of your comments even farther than you did...

That is the reason for the S-curve--it increases tonal range in the mid range, often at the cost of tonal range elsewhere (depending on what the total range in the image is, and how the midrange is expanded).

If there are three ranges of luminosity in the image -- shadows, mid tones and highlights -- it's actually impossible to apply a traditional S-curve to increase the mid-tone contrast without decreasing the contrast in the shadows and highlights.

However, an important point needs clarifying...

The tonal range has been stretched out considerably. This was done with the LR contrast slider, but I would have obtained a very similar histogram by using the blacks and white sliders, which among other things change the black and white points. I would expect the image to show more contrast. And indeed it does

The two histograms you displayed prove that the Lightroom contrast slider moved the black and white points. As you know, doing so inherently adjusts all other parts of the tone curve; it's impossible to adjust the black and white points without adjusting the rest of the tone curve. I suspect that the contrast slider makes additional adjustments to the mid-tones area of the tone curve as well, much as what would happen if a traditional S-curve was applied. People less informed than you use such sliders not having any idea about what is really happening to the tone curve.

[AlwaysOnAuto]

Mike, and Dan, which of the images Dan posted do you think really represents what it actually looked like when that photo was taken?
Personally, I don't think the second, edited one, looks 'natural' or real. I feel the edit made it too 'idealized' for my tastes. I guess for my own photography I want my photos to represent the actual 'moment captured', not an 'idealized' version, even if it doesn't look as 'pretty' to others.

[DanK]

The two histograms you displayed prove that the Lightroom contrast slider moves the black and white points. As you know, doing so inherently adjusts all other parts of the tone curve; it's impossible to adjust the black and white points without adjusting the rest of the tone curve. I suspect that the contrast slider makes additional adjustments to the mid-tone area of the tone curve as well, much as what would happen if a traditional S-curve had been applied. People less informed than you use such sliders not having any idea about what is really happening to the tone curve.

True, but I think only partly on track for this discussion. The important point, IMHO, is your point that stretching the histogram by adjusting the white and black points will alter the rest of the distribution as well. The fact that the lightroom sliders do more than that is certainly true, but I don't think it matters for the discussion. Here is another image, this time created by taking the low-contrast image above and using a levels adjustment in photoshop to change the black and white points. It is not as extreme as the LR image, but it too is much higher contrast than the original.



I'm really only trying to make a couple of points. First, the histogram contains all information about the frequency distribution of luminance, so it contains information about contrast, and adjustments to contrast, however one chooses to make them, will alter the histogram. Second, in some cases, you can interpret the histogram directly for that purpose. For example, a narrow histogram, with a high black point and low white point, will almost always be a low-contrast image, and stretching it will increase contrast.

AOA, I'm not making an argument for one or another kind of edit. That's for each photographer to decide. I was just picking up on a comment that John made and trying to clarify the relationships between the histogram and the perception of contrast.

[Manfred M]

Mike, and Dan, which of the images Dan posted do you think really represents what it actually looked like when that photo was taken?
Personally, I don't think the second, edited one, looks 'natural' or real. I feel the edit made it too 'idealized' for my tastes. I guess for my own photography I want my photos to represent the actual 'moment captured', not an 'idealized' version, even if it doesn't look as 'pretty' to others.

I think we could have a very long and interesting discussion on what the eye sees and what we see on our screens and prints.

Let's remember that, depending on your source, the human eye can perceive a dynamic range of somewhere between 12 and 20 EV. There does not seem to be a particularly clear consensus on this. A modern medium range camera can capture around 12 EV of dynamic range. A paper print is can reproduce around 4 EV and a decent computer screen is in the 5.5 to 6 EV range.

Your eyes / brain (and camera) can capture far more dynamic range than any of the output devices that we use in photography. The means colours tend to be more brilliant in real life than any photographic image. I suspect Dan's second image is probably closer to real life than the first.

In order to display our 12 EV image on a computer screen or print, our cameras (in producing jpegs), computer screens and printer drivers rearrange (i.e. compresses) the data so that it can be displayed. Some software engineer has decided an "optimal" way of displaying the data, and this often results in an image that really does not look at all like what we saw. We've looked at so many jpeg image, we might be fooled into thinking this is what the scene should have looked like. Most often, the black point (i.e. the colour that is "pure black" in our image, i.e. the darkest shadow detail is in fact rendered as a gray value. Pure whites have exactly the same issue and are also often rendered as a shade of gray. Ensuring that blacks are black and whites are white more closely matches the way we see, so in fact, in my opinion, provides a "more natural" looking image.

We have the same issue in shots where we see blue in the sky, but our images show gray or white without a hint of blue. Again, the compression algorithms have not done a good job in reproducing what our eyes see, so pulling out colours that the automated algorithms haven't actually produces a more realistic image, not one that is less so.

Carefully said, yes, one can go overboard in post-processing and create images that do not look realistic. I rarely use the saturation slider to increase saturation, because the colours will not look natural. On the other hand, if I shoot the sky with a polarizing filter, I can create a really funky, unrealistic looking sky, based on how the camera processes the image, so care has to be taken when shooting skies; I tend to maximize the impact of the polarizer on the scene, but then will dial it back a bit to ensure that my skies don't look strange.

The colours that you in real life are much more brilliant than the colours we normally see on our computer screens (transmitted light, additive RGB colours), and these in turn are more brilliant that we see in our prints (reflected light, subractive CMYK colours).

[Mike Buckley]

Mike, and Dan, which of the images Dan posted do you think really represents what it actually looked like when that photo was taken?

I wasn't there, so I have no idea. However, Dan lives in New England. If he took that photo in New England, the second photo could easily represent what it actually looked like more than the first one. It looks both natural and real to me based on typical New England vistas I have experienced.

I guess for my own photography I want my photos to represent the actual 'moment captured', not an 'idealized' version, even if it doesn't look as 'pretty' to others.

There's not a thing wrong with that. Just be aware of the limitations of doing that, as explained by Manfred.

[AlwaysOnAuto]

I really appreciate all the input given here. It's obvious to me that I have a lot to learn when it comes to effective PP of my photos. Thank you all for taking the time to present your views/ideas/thoughts.

[ajohnw]

The histogram can have another use. When I looked at Dan's histogram I had the feeling from it that the colour balance was some what askew due to the rgb distribution. The GIMP has an auto levels that alters the usual 3 pointers on each channel individually. This is what that does in this case.



It's made significant changes to the span of all channel inputs. Rather a lot on blue and left luminance alone. In effect it's trying to correct the RGB response of the camera based on the shape of each channels histogram. sometimes it's a disaster and not easy to do manually however the position of the peaks in the histogram and the total span of each channel can give an indication of what to do.

I tried to steer the OP away from curves as people seem to find them difficult. Recovery sliders are easier and minor changes to a curve can make a considerable difference to the image. Levels is much easier to use. It's effectively generating an S curve. I'd hope a CinC tutorial explains what each pointer does. It might mention gamma but it's better to read that as mid tone contrast. Gamma is a subject on it's own applied for very specific purposes. Using the term here just adds to confusion.

Curves are pretty simple really. The tonal range of the image going in is on the vertical axis and what will go out on the horizontal one. Increasing the slope of the line increases contrast in that region, decreasing it has the opposite effect. CinC tutorials should clear up the rest.

John
-

[Mike Buckley]

For example, a narrow histogram, with a high black point and low white point, will almost always be a low-contrast image, and stretching it will increase contrast.

That's true. As I woke up this morning and thought about it, I realized that I had ignored that particular circumstance. I then realized that that's because when the histogram is so narrow, all but the absolute very beginner would see the lack of contrast in the image without the aid of the histogram. My bad.

[Mike Buckley]

Curves are pretty simple really. The tonal range of the image going in is on the vertical axis and what will go out on the horizontal one.

That detail isn't accurate, John, at least not with any Curves tool I have used. The Curves tool is a graph. Any point on the graph is where the pertinent point of the horizontal axis meets the pertinent point on the vertical axis. The straight line of the tool is the beginning mapping of those points; some if not all software allows you to hover your pointer over any tone in the image and the corresponding point on that straight line will be indicated. Once you alter the tone curve, you are mapping the luminosity on the straight line to the revised luminosity on the altered curve.

EDIT: I just now realized that we're probably saying the same thing using different words. Even so, on the Curve tool that I use, the input value is on the horizontal axis and the output value is on the vertical axis (not the other way around as you describe). Considering that there are 256 points on both axis, it doesn't matter which axis is the input and which axis is the output.

[ajohnw]

That detail isn't accurate, John, at least not with any Curves tool I have used. The Curves tool is a graph. Any point on the graph is where the pertinent point of the horizontal axis meets the pertinent point on the vertical axis. The straight line of the tool is the beginning mapping of those points; some if not all software allows you to hover your pointer over any tone in the image and the corresponding point on that straight line will be indicated. Once you alter the tone curve, you are mapping the luminosity on the straight line to the revised luminosity on the altered curve.

The steepness of the line is pertinent which is how I looked at it from day 1 as that was very apparent. GIMP which gives no clue at all. The initial straight line going from corner represents 1:1 input to output. A typical S curve reduces contrast at the dark and light end and boost in between the control points. An inverted S does the opposite. Have I reversed what the axis represent? I might well do and leave people to think about that.

True some packages will show a line on the curves plot indicating the tone level under the mouse cursor. Some will also add a luminosity histogram to the curve view, also in some case bounding - a sort of indication how much the curve can be altered. I've never found that of much use. Some packages might offer all 3 and more. I understand that some Adobe products offer sliders which manipulate the curves. Fotoxx does too - contrast and brightness. Increasing contrast generates an s curve, decreasing it an inverse S, brightness moves mid tones up. All of these are describing what happens to mid tones.

I didn't mention brightness but taking a simple example, moving the extreme black end to the right, this will darken that end and boost the contrast past the point that was moved. This is effectively moving the black point. Moving that point up will lighten it ( Indicating that I probably did get the axis the wrong way round), slope will reduce though so contrast is reduced.



As it takes longer to upload them a few examples might help - not aimed at good images just showing principles. Hope Dan doesn't mind me using his image.

S curve


Inverse S


A mix


Best avoided but note the increase in the reflections bottom right. Really this is a case where a package that has selection should be used and a separate curve applied to the selected area or a curve brushed onto it.


That sort of thing is how a mess can easily be introduced.

John
-

[ajohnw]

There is one other form of curve that some might come across that maps as Mike puts it the raw file tone range into the output tone range. This can be a bit weird to use as getting all of the bright end of the raw file in may well mean dragging the bright end down a lot. The behaviour isn't the same when this arrangement is used, but similar.. Some raw conversion packages may have this and are also likely to have a normal luminosity one as well.

Some can also show a raw histogram. Some people look at that and say woopee the luminosity range is ok - or drat or what ever if it is too large. Too much colour channel clipping can be seen at this level as well.

John
-

[DanK]

As to which is more like what we saw: that is a far more difficult question.

First, it looked nothing like what John got from Gimp's automatic correction in post 49. My guess is that Gimp (which I don't use) is trying to balance the three channels, but there is no reason to expect them to be balanced in real life.

Second, as for my first two in post 40: neither one is really on target. This is complex, at least for me, and I handle shots like this more poorly than almost any other. For those of us there at the time--and there were quite a few also trying to photograph--the scene was beautiful. People would just stand and stare. The terrain is rugged and majestic, and the colors underneath the fog were at their peak. So, I experienced it as a beautiful scene masked by a mist. Take a picture, and you get the first of the images in post #40: washed out, lacking dynamic range and hence contrast, and lacking saturation. Boring, in a word. AoA (please add your name to your profile--we use real names here), you might want this because it is SOOC, but for my tastes, it is a dull snapshot to which no one would pay much attention.

Some of the problem may be the technical issue that Manfred described, but I think that it more than that. I think it is more than our brains adjusting what we see. It is also a matter of what we perceive. In the real world, surrounded by this scenery, we process it for all of what I described. Reduce it to a small 2-dimensional image, and we don't.

What to do? A common strategy for dealing with mist is to frame a shot so that something is not obscured and is in crisp focus, for contrast. No chance of that here. So, I was left trying to figure out how to process the image to convey some of the grandeur of the scene.

I posted a bunch of these at the time, and I realized after the fact that the approach I followed was probably not a good one, in one respect. I increased both the luminance range and the contrast in certain ranges. When you do that with the most common methods, you are also boosting saturation, and once that reaches a certain point, it begins to look artificial.

I think the best approach might be to edit the shots in LAB in order to separate luminance from saturation. I have never done that, but I think I will try it with a few of these. The alternative, which I have used a number of times, is to apply a curves adjustment with the blend mode set to luminance. That works fine, as long as the curves adjustment is the only major adjustment to tonality, which was not the case here. You can do this in Photoshop, but I don't think it is available in LR yet.

This is a bit of a digression, but the LR and photoshop (without a luminance blend mode) curve tools affect color differently. I have stumbled on readings about this but haven't taken the time yet to master this. One reading is here.