Kind of unusual level/curves question ...

[dmr]

I do consider myself fairly fluent in Photoshop, but today I ran across something I haven't seen before, or at least don't remember seeing it.

I was editing a scan I did last night, and the image looked quite normal, and as a part of the usual tweaking I went into levels to see if they looked sane.

Well, there were no significant flat portions at the darkest and lightest parts of the histogram. I'm used to these, and used to correcting them.

However, right DEAD CENTER is a significant flat portion, which indicates to me that for some reason in this photo there is a lack of significant information right in what should be the middle tones.

My instinct told me to move the mid-tone slider to the left, but I was wondering, is there a way that anybody knows to maybe compress that flat section, to make use of the entire tonal range with what information is in the photo.

I'm attaching a capture of the levels histogram, and of the photo itself. The photo looks normal, but the histogram is unusual.

Comments?

Thanks, gang.

Bonus question: What film is this?

 

[Pherdinand]

I'm much less fluent in PS but here's what I think:
The histogram is scaled vertically, i mean, the midtone info is there but it's hardly visible due to the large bumps on the sides.
You have a large bright area which is very uniform, the sky. You also have alot of the same intensity on the dark side, even if it is well distributed. I think you REALLY have much more of the extremities than in the middle region.
Solution? Dunno, maybe playing with the blue histogram only? the image is too blue anyway, i'd say, and you might reduce the two bumps on the sides by messing around with the blue only. But why do you want to have a "sane histogram" if the image looks normal?

EDIT: is this the 8 or 16-bit histogram?

 

[Robin Harrison]

True - this isn't common, but there's nothing 'wrong' with a histogram like this. Yes - you could compress this section by making a curves adjustment with a 'step' in the midtone section, but you have to ask yourself why would you do this? The histogram is a useful aid, but adjustments shouldn't be made just to make it look 'normal'. The effect of that curves adjustment would be to move 40% tones and 60% tones only about 5% apart. You would reduce the seperation in the midtones and my guess would be the picture would look awful as a result.

 

[dmr]

is this the 8 or 16-bit histogram?

The histogram is before I converted to 8 bit, which in this version (7) you have to in order to do many things.

In Photoshop 7, I'm really not sure what functions are true 16 bit and what are 8 bit. If the histogram were a true 16 bit, I would assume the levels would go from 0 to 65,000 and something, but they are noted as 0 thru 255.

 

[Pherdinand]

If the histogram were a true 16 bit, I would assume the levels would go from 0 to 65,000 and something, but they are noted as 0 thru 255.

That's indeed strange; however, when you do changes in the histogram in 16-bit version, the result indicates that it IS a 16-bit operation. E.g. when you simply expand it in 8-bit by shifting the high and low sliders, you get gaps in the histogram, which you do NOT get in 16-bit.

 

[dmr]

however, when you do changes in the histogram in 16-bit version, the result indicates that it IS a 16-bit operation. E.g. when you simply expand it in 8-bit by shifting the high and low sliders, you get gaps in the histogram, which you do NOT get in 16-bit.

And this is why I try to do levels and such before switching to 8 bit.

I've been playing around a bit. I made 2 copies of this, one where I expanded the high tones to fill the histogram range, and another where I expanded the low tones to fill the range. Then I converted both to 8 bit, put them in different layers and played around with the opacity.

Then, as you suggested, I tweaked the midpoint of the blue histogram. What I am attaching here is closer to what I want, which shows more detail in the clouds.

I'm still playing ... thanks everybody.

 

[JohnL]

My suggestion would be to see what the effect is of raising (edit: lowering, oops) the curve in the 90% region, and dropping (edit: raising) it in the 20% region while holding it in the 10% region. This should give you a steeper (edit: shallower) slope in the (almost non-existent) upper-mid tones. Of course, the numbers above are just guesses. You'd have to play around with it. Something like this giving a result something like this .
It looks a bit blue but that is fixable.

 

[jlw]

I do consider myself fairly fluent in Photoshop, but today I ran across something I haven't seen before, or at least don't remember seeing it...

May seem like an insultingly obvious question, but it's coming from someone who's been there --

-- You didn't accidentally bring up Levels while you had something selected in the image, did you?

If you've hidden the selection borders (command-H) it's embarrassingly easy to get into all kinds of weird head-scratching situations before you realize you're working only on a selection, not the whole image.

 

[Gabriel M.A.]

I don't think your question is unusual, don't worry.

What's going on here is that you have more information in the shadows and the highlights, that's all; the camera meter obviously got fooled when averaging the overall scene. Had you metered the ground first and then taken the shot, you may have had a blown-out sky, but the buildings would have been brighter, in the midtones, and you would have had more information "in the middle" of your histogram. Does that make sense?

 

[dmr]

You didn't accidentally bring up Levels while you had something selected in the image, did you?

No, I was going thru more or less a standard sequence of what I do when I have a fresh scan. I just checked the levels and noticed the rather unusual histogram.

Had you metered the ground first and then taken the shot, you may have had a blown-out sky, but the buildings would have been brighter, in the midtones, and you would have had more information "in the middle" of your histogram. Does that make sense?

Uh-huh, sure does. And, I'm glad I didn't meter downward, as I'm kind of in the habit of doing, since I do want the detail of the clouds in there.

Thanks everybody, I appreciate the advice and the comments.