Shooting quilts with flash indoors

[ajohnw]

Looks like more than enough to go on with to me. The controls in PP packages are often organised in the order they should be used. Some make major changes to get the image right and other augment it.

The key really is the histogram top left. Dark end to the left bright end to the right. The jpg you posted initially is a good one for getting a grip on how that changes as the histogram is a hump that doesn't take up all of the "space". I ran the dark end very slightly of the end of the histogram - referred to as clipping. The black point moves that end around - sort of fine exposure compensation and allows clipping to be set more precisely. If you work on the jpg set that to zero and see what the adjustments I mentioned do. When a shot is reduced I like to save it and load it again to sharpen.

Googling the adjustment names can help too as most PP packages have the same ones. With luck links can be found that actually state and show what they do.

John
-

[AlwaysOnAuto]

When I read thru the tutorial on histograms, I kept looking for something that would tell me what the 'ideal' histogram would/should look like, but I couldn't find anything.
Is there such a thing?
Or is it kind of like the tuning of an Alfa Romeo SPICA injection unit thermal actuator? Each one has its own characteristics and the 'factory spec' is just a good starting point for adjustment, but is only that, a starting point. Each one must be fine tuned to its particular characteristics. Once I had that epiphany I was able to dial it in no problem.

[Mike Buckley]

There is no ideal histogram. That's because all images are at least somewhat different and so should their corresponding histograms.

However, there are histograms that routinely can be improved upon by changing the exposure, the tone curve and the black and white points. Even then, it's the improved characteristics of the photo that matters, not the histogram; the histogram is merely a graphical representation of the data and as such can be a helpful guide that makes it apparent how you could perhaps improve the image.

[AlwaysOnAuto]

I've been playing around today with the file and got a version where you can actually see the fold/bend of the quilt which isn't hanging real flat. I think you version John is so bright it doesn't actually show this 'feature' if you will.



This a reduced image of it since it was a 19.3mb file to start with. I haven't done any PP on this reduced image as a reduced image.

[DanK]

When I read thru the tutorial on histograms, I kept looking for something that would tell me what the 'ideal' histogram would/should look like, but I couldn't find anything.
Is there such a thing?
Or is it kind of like the tuning of an Alfa Romeo SPICA injection unit thermal actuator? Each one has its own characteristics and the 'factory spec' is just a good starting point for adjustment, but is only that, a starting point. Each one must be fine tuned to its particular characteristics. Once I had that epiphany I was able to dial it in no problem.

Mike's answer to this is very good. I'll just add a bit. In the example you give, each instance of the machinery is a bit different, but the desired end point is the same. Mike's point is that when editing images, the end points are not the same either. For example, you may have one image that you think should have a lot of contrast and another that shouldn't. Or to take another, related example: a bunched up histogram represents a limited range of luminance. Sometimes that is exactly what you want. For other images, it is what you don't want. The starting point should be: what do I want this image to look like? Then, as Mike said, the histogram gives you a display of the data that helps you get to where you want to go.

[AlwaysOnAuto]

Thanks Dan, Mike and John.
I guess I need to cure my ignorance of how the histogram relates to the images and where I want to go with them. I'm a complete novice at this PP stuff so it's gonna take some time.

[ajohnw]

The histogram is a guide really to what can be done. The jpg is an example with little luminance range so it's histogram takes up little of the available "space" - that is the entire size of it's view. When working on other shots from raw the file can easily have more range in it than can fit into the histogram. Keeping things simple this is what recovery sliders are for - usually marked shadow and highlight or maybe fill light in respect to dark areas. These will move the raw files tone range that isn't visible in the histogram into it - often a good place to start is to get the lot in. Exposure compensation will also shift the histogram around. Best thing to do is to play around with the controls on several shots as they all inter react.

The next adjustment again trying to keep it simple is to adjust gamma - mid range contrast after a fashion - the easiest way to do this is levels which I think you have. It should have a centre pointer - move it around delicately and see what happens.

In practice shots sometimes have to have some clipping either in the high low or both lights. Some shots can even look better clipped some what as per the black in the shot I posted. You'll probably have to accept for a while that doing this leaves more room for contrast in the rest of the shot.

Many people use an adjustment called curves rather than levels also to perform some of the other aspects such as high and low light tone behaviour. It may pay to leave that till later in your learning curve.

On your too bright comments I suspect you have another problem - display settings. Monitors these days are often too bright straight out of the box. Photographers use specific brightness levels. Given that monitor manufacturers don't set to these by default I can't help wondering what will happen over time. The best answer to this problem is more expense - a colorimeter however the CinC tutorials contain one on setting a monitor up roughly without one. There is also this page that may include your monitor, disregard any that don't give monitor settings.

http://www.tftcentral.co.uk/articles/icc_profiles.htm

If your monitor is there you should also install the ICC file on your machine as well as using the monitor settings.

John
-

[Manfred M]

Thanks Dan, Mike and John.
I guess I need to cure my ignorance of how the histogram relates to the images and where I want to go with them. I'm a complete novice at this PP stuff so it's gonna take some time.

We were all novices at some point.

In general, getting it right in the camera is going to save you a lot of work in the long run; as a rule of thumb I find that in general an extra minute spent taking a picture saves at least 10 minutes fiddling in post-processing. Fiddling with gradients to compensate for uneven lighting can be rather fiddlely.

Getting the proper exposure is of course very important and if you are not certain, try bracketing your shots. By the way, there is no "ideal" historgram; a histogram merely shows you the colour or tonal distribution of the image. The positioning of the histogram is more important; if all of the data seems crammed on the left side, this tends to indicate underexposure and if it is towards the right, overexposure (but it can also mean a predominance of very dark or very light tones too).

With your previous comment regarding straight out of camera (SOOC) images, again, one has to be very careful there too. If you think of a modern camera as a computer that takes pictures (which it really is), then you are relying of some engineer or computer programmer's view as to what your image should look like. In the case of an average scene (whatever that may be), perhaps this is not a bad assumption, but I'm pretty sure that they never considered images of a quilt when they were designing the camera's processing algorithms.

[AlwaysOnAuto]

Your last paragraph is very true Manfred, but I've also accepted the fact that their guesses are a lot better than mine are what with my lack of experience in taking pictures. Thanks for your input.
Now on to monitor calibration...

[Mike Buckley]

I've also accepted the fact that their guesses are a lot better than mine are what with my lack of experience in taking pictures.

That's understandable so long as you achieve the ideal exposure. The way to do that is to check the histogram after capturing the image. If an adjustment is needed, you can then retake the photo and check the histogram again. Repeat that process until your camera displays a histogram that is appropriate for the scene being photographed. Naturally, all of that requires understanding what the histogram indicates (and doesn't indicate) and how you can use it to determine the ideal exposure.

[AlwaysOnAuto]

Your last sentence says it all Mike. Thanks for taking the time to help.

[DanK]

There are good tutorials to help you understand histograms here: https://www.cambridgeincolour.com/ph...-tutorials.htm. The x-axis ranges from fully dark at the left end to maximum brightness at the right end. The height of the histogram represents the density of pixels at that brightness. You can get them showing total luminance (gray in most software I have used) or to show the luminance of each color channel separately.

However, once you understand that, you have to learn how it relates to the appearance of the image, as it is and as you want it to be. I'll use two images to illustrate.

First, a landscape:



This is a well-illuminated shot with a full range of luminance, from light to dark. The histogram shows this full range, as well as a peak around the middle, which is fine for this type of shot:



Now, look at this shot of a milkweed taken against a black background:



Because much of the frame is fully black, there should be a concentration of pixels at the left edge. And because the filaments go to almost pure white, a thin piece of the histogram should extend almost all the way to the right. And that is exactly what it looks like:



If I edited the second shot to give me a histogram like the first, it would be a mess. In fact, I did the opposite: I edited it to force the background to true black, pushing the peak in the histogram even farther to the left.

In most cases, you won't want a histogram like the second. In general, any area in which you want detail should not be so poorly exposed, because it will increase the appearance of noise. However, this pair illustrates that there is no one ideal histogram.

[AlwaysOnAuto]

Thanks Dan, very helpful post.

[ajohnw]

Dan's landscape shot shows another important point - slight clipping of blacks but none of highlights. This often crops up, cases where the highlights are most important. One thing that people sometimes find confusing is that when we look at a scene like that we don't usually see any shadows. The camera does. If we can actually see them the camera is very likely to strengthen them alarmingly. Dan has probably lifted them is this shot so that they blend in with the image and only small areas are clipped. The other solution to the problem is to only take shots when lighting is suitable.

The 2nd one also shows that the white stops well short of full white but these parts of the image still look white. The same can apply to blacks. To look black they needn't be full black. Again academic but can give an indication of what can be done to an image.

Academic? The final result really should be what the person who is doing it wants to achieve.

John
-

[DanK]

The height of the peaks give some idea of where the maximum levels of contrast are. If for instance the highlight peaks were lower the clouds would look flatter.

John,

I don't think that is right. If an area of the histogram is pushed lower, that simply means that there are fewer pixels at that level of brightness. Contrast is the difference in brightness, not the amount of pixels at one level. If you had an image with exactly one tone, the entire histogram would be just a spike at one value--the maximum height--but there would be no contrast whatever.

In general, contrast is increased when the histogram is stretched on the x axis. Take a low-contrast image, increase the contrast using any software that has a contrast adjustment, and see what happens to the histogram. Or take an image that has a limited range of luminance, and change the black and white points to increase the spread. The result is more contrast. A traditional S-curve on a curves adjustment, if I am thinking about this right, simply compresses both extremes, increasing the appearance of contrast by increasing the luminance range in the midtones.

Or am I missing something?

Dan

[ajohnw]

Perfectly correct Dan. I shouldn't have mentioned something that I don't actually use the histogram for. Come from thinking about it as typing it and forgetting luminosity goes across not up. I've edited the post. Thanks for pointing this out.

John
-

[Mike Buckley]

In general, contrast is increased when the histogram is stretched on the x axis.

That depends on the meaning of the term, "contrast," to you. I see a lot of people using the meaning that way. However, for me, that part of the histogram has to do with tonal range, not contrast. If an image has a lot of pure whites and pure blacks it will have maximum tonal range. Even so, it can be a low-contrast image, at least as the term, "contrast," means to me.

That's because when I think of contrast, I automatically think of mid-tone contrast. That's the range of luminosity where most contrast is generally affected.

Indeed, the histogram doesn't convey any actionable information to me pertaining to contrast. All of the actionable information has to with tonal range, clipped shadows and clipped highlights.

A traditional S-curve on a curves adjustment, if I am thinking about this right, simply compresses both extremes, increasing the appearance of contrast by increasing the luminance range in the midtones.

A traditional S-curve doesn't affect the extremes; the black and white points remain the same. You're correct that it increases the luminance range of the mid-tones by remapping the darker mid-tones to even darker tones and remapping brighter mid-tones to even brighter tones. More important, doing so inherently increases the mid-tone contrast because the difference in luminosity of adjacent tones after applying the S-Curve is greater (more contrasty). As a purely hypothetical example, assume the mid-tone luminosity values of 127 and 128 in an image are re-mapped using an S-curve to 125 and 130. As you see, the adjacent tones of 125 and 130 will display a lot more contrast than the adjacent tones of 127 and 128. That explains why the relatively vertical part of the S-curve encompasses the mid-tones and the area of increased contrast.

[Mike Buckley]

It just occurred to me that Manfred recently posted two images of the Taj Mahal. When I take his low-contrast image and adjust the white and black points to be the same as his higher-contrast image, the revised image still has very little contrast. That's despite that the difference between the brightest and darkest tones is greater, which people often call increased contrast. Indeed, the only way I can alter his low-contrast image to display the contrast of his higher-contrast image is to then increase the mid-tone contrast by applying a fairly aggressive S-curve.

[AlwaysOnAuto]

Interesting discussion gentlemen, but, the only S-curves I can really relate to are/were the esses at the old Riverside Raceway.

Carry on.

[DanK]

Mike,

I partly agree with you, but I disagree with some things you wrote. I'm not sure how much of it is just semantics, i.e., how we use the word "contrast."

That's because when I think of contrast, I automatically think of mid-tone contrast. That's the range of luminosity where most contrast is generally affected.

Yes, tonal range in the mid-range is what is often perceived as contrast. 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).

A traditional S-curve doesn't affect the extremes; the black and white points remain the same.

Half right. It doesn't affect the limits of the histogram on the x-axis, that is, the black and white points. But it does do what I wrote, which is compress the extremes. If the black and white points remain fixed, an S-curve necessarily compresses the extremes. There is a fixed range of luminance. If you take up more of that range for the midtones, you leave less for the extremes. The flat parts of the s-curve are where you are compressing the range. That is typically in the extremes. In effect, the S-curve transfers some of the luminance range from the extremes--the regions where the slope is less than the original diagonal--to the midtones, where the slope is now steeper.

Indeed, the histogram doesn't convey any actionable information to me pertaining to contrast. All of the actionable information has to with tonal range, clipped shadows and clipped highlights.

The information is there, and it is often actionable. All information about the density distribution of luminance is displayed in the histogram, so anything that depends on that distribution, including contrast, has to be reflected in it. Sometimes it is hard to decipher this from a histogram, but sometimes it isn't. I'll make this concrete. Take a look at this histogram:



This tells me that the image has limited tonal range and is likely to look low in contrast. And indeed it does:



Now compare this histogram, from a simple edit of that image:



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:



Of course, by stretching the histogram, this also stretched the midrange, because in this case there was room to stretch it all.

Where I think we get to semantics, or at least to subjective judgment, is what people perceive as contrast. There is no doubt that people generally perceive a stretching of the midtones as contrast. I also often perceive the total width of the distribution, e.g., the presence of true blacks and true whites, or tones close to them. as affecting my perception of contrast.

In practice, I often use the width of the histogram this way. One can also use portions of it to help adjust contrast.

Dan