Tony,
The middle slider is a pure gamma (power) function between levels 26-255. It transitions from there down to level 16 and finishes as a linear function from level 16 to 0.
This is done to avoid an infinite slope of the input/outout curve at zero input, which a pure power function produces.
To say that the modification "damages" anything implies that the system/monitor response in the deep shadows is well behaved and itself follows a pure power curve—and nobody knows that to be true.
For the levels middle slider (or any of the other tools, for that matter) to correct system/monitor non-linearities, the non-linear behavior being corrected needs to be known. In the deep shadows it is not known. So one extrapolation is as good as another.
George
Hello George
There are other gamma utilities that do not do slope limiting. The result is a much more open shadow than that provided by Photoshop. This openness also shows the shadow noise in scanned images. Perhaps Adobe have done the limiting because the effect not noticable on the final print.
It is this clipping in Photoshop that I do not like. I would rather not have it. I can always do the clipping myself.
I feel that a more useful tool would be to do all gamma in the curves dialogue with a middle input box and a histogram.
Mike Engles
Hi Mike,
I hear you. I am thinking that it doesn’t make sense to fuss over slope details in the deep shadows for an 8-bit system which can’t track refinements anyway.
George
I agree. Since installing the Epson 3200 and using the Epson software, their gamma slider does a much better job on open shadows. In PS, I always check between curves and gamma slider to see which works best. Now, I don’t need to generally export to PS from the scanner in 16 bit.
Mike, to what gamma utilities are you referring? I don’t believe any gamma slope correction can be implemented without normalization.
Hello
I suppose that I was being misleading about utilities. I meant the Gamma curve function in Picture Publisher as well as the AIM plugin for Photoshop and Nikonscan twain. They work without the limiting.
Mike Engles
I haven’t run into problems but then I seldom touch the middle slider in LEVELS. I simply adjust the end points by minimizing the part off the toe of the histogram. I let the middle slider ride. The rest of the adjusting I FINALLY am able to do with CURVES.
I scan in Nikon LS-40 (or that for the Epson 2450 depending) color space and edit in those too. With my Epson 2200 printer, 16 bit color and those color spaces; files do print better than the same file in Adobe RGB.
I think the next test will be interesting. The Epsons (current and the now retired 1270) give a slightly better print at 360 dpi file resolution. With the 2200 having twice the resolution of the 1270, I’m going to spend the time and ink to see just what happens at a file resolution of 720 dpi at full printer resolution. I’m already planning to get my hair cut while it prints that monster. <grin>
As for Nikonscan twain, try VueScan, www.hamrick.com. I found it to be FAR better than Nikonscan. It does have a learning curve and you will have to adjust the curves, but there is more picture information in a VueScan file than a Nikonscan file. I did several different scans of the same negative with each, worked them up in Photoshop, and in every case, the best I could do with a Nikonscan source image wasn’t close to meeting the quality of the worked up VueScan one.
Bill
So far as I know, the 2450 exports to PS on Adobe RGB. Are you using the Epson Twain or the SilverFast?
I still don’t understand how a non-linear function can operate without some limiting. Otherwise it "blows up" to infinity.
I just bought Photo Elements 2 and I find there is no toolbar for making gama correction. Feedback as to how to do this would be appreciated
Thanks
Mak Mahesh
Hello
I have used the demo of Vuescan since it first came out. Initially I found it very good and was pleased with the result, so I nearly bought it. I then noticed all sorts of strange colours and artifacts. I came to the conclusion that NikonScan was ‘truer’. I have tried every version since, but alwys prefer NikonScan. I have never actually bought it.
Mike Engles
I was not too thrilled with Vuescan either.
Native software better than third party. What is this world comming to?
Tony:
Any mention of Timo’s site on this forum will result in a good tongue-lashing from Chris Cox. Surprised Len didn’t jump up and down.
There is one valuable aspect to the site though. How to set your monitor up correctly by setting the black point, that is correct.
The rest of the site deals with a difference of opinion, the only problem being Timo’s ongoing attacks of other individuals (Charles Poynton, Bruce Fraser) of impeccable qualifications in the field who can explain their point of view.
Timo thinks gamma should be linear to be perfectly correct (I actually scan without a profile and use a linear gamma), the other camp says no, there has to be compensation with hardware colour profiles and non-linear gamma settings to accurately represent what the human eye sees. The argument is more complex than that, but that is the simple jest of it.
How did we get from Gamma slider to system gamma? These arguments have been hashed and re-hashed over and over.
All this stuff about system gamma should not be even discussed here. It is the privy of the HW designers to deal with it. Crt’s are governed by power laws. The designer of that system is responsible to provide necessary corrections so that a linear input provides an approximate linear output to the eye, likre audio volume controls. 0 to 100 IRE (NTSC) are linear steps from total black to total white, or nearly so. That nearly so is what the designer has to reconcile.
I don’t want to even know about it. Just tell me what voltages it needs.
Norman Koren does some of the best prints I have seen. I spent a couple of days last fall with him when he visited his son in Portland.
He doesn’t use Photoshop, but he didn’t bad mouth it to me either.
Power laws can get out of hand, especially when they are applied sequentially.
How did we get from Gamma slider to system gamma?
As I read it, Timo says the gamma slider is inaccurate, and his argument is that all monitors gamma should be set to 1. I think.<grin>
I was asking rhetorically Tony. Wasn’t aimed at you. Sometimes I get too caught up. Must curb those tendencies.
Peace,
Larry
Sometimes I get too caught up. Must curb those tendencies.
Me too Lawrence. If I’d had my wits about me, I probably would have realized that it was rhetorical. Unfortunately I don’t <grin>
Peace,
Tony
Ah, That one is extra good. 🙂
Simple: levels are linear weighting, curves are non-linear weighting.
Congratulations Tony – you just stumbled onto a well known mis-information site.
Timo is well known for trolling newsgroups and spreading misinformation.
Anything you "learned" on that site, is probably wrong.
Am I right or am I right? Huh, huh?
Tony – yeah, I saw your comments. But I just HAD to fulfull the prophecy…. LOL!
I’ve seen Timo’s site before. Wheter he was right or not, weighing the information with a "usefulness" filter, I quickly left it behind.
Speaking of weighting, I don’t understand what linear and non-linear weighting mean here. Linear weighting, if it is linear, either changes the angle of the slope by some selected value, or introduces an offset. Non-linear weighting can mean almost anything. If these are particular terms to PS, I would like a definition.
"…What is the difference between the algorithms in the levels slider, and the ones governing Curves?…"
Lawrence, The Curves tool does not apply an algorithm. That’s why it is able to modify output as a function of input to achieve ANY effect, including the effects attainable with other tools (such as Levels) which DO use algorithms and are constrained thereby. With Curves, YOU become the designer of the effect—not just a parameter adjuster. That’s why Curves is so powerful.
ID Awe, "…Simple: levels are linear weighting, curves are non-linear weighting…" Sorry, Curves can be linear or non-linear—as you desire, and Levels sometimes is linear and sometimes non-linear, depending on which slider you use.
George
That makes perfect sense, George. With a little reflection, it’s easy to see.
I am the algorithm!
Thanks.
"…George: Disagree. Think about that again…"
ID,
Been unable to think. Your cryptic message (above) has caused the decades-old hit-song "Is that all there is?" to resurface, and I can’t expunge it from my consciousness. How diabolical can you get? (-:
George
Told you that you were a sadist <grin>.
Interesting that the calibration of the middle slider should be the reciprocal of the exponent. Perhaps because of our expectations that lower is darker.
Lawrence,
The reciprocal is used because the slider value then becomes not the power being applied but the power you are correcting for. To linearize a power curve with power 2.2, for example, you enter 2.2 in the slider box and 1/2.2 is applied to the input, cancelling the existing non-linear 2.2 power input, if that is your intent.
George
Yes, I see that. What I am wondering is why that information is important to a user. The slider could be 0 to 100 so far as I am concerned, as a user. Or 0 +/- 50, say.
BTW, the Epson 3200 scanner now shows a histogram and a transfer function in it’s Levels dialog box, called a Tone Curve Viewer. What is intersting about it is the way the Gamma Slider (calibrated in reciprocals also) changes the curve shape. It seldom (perhaps never) linearizes the Tone Curve. It attempts to, but as you approach the higher values, a curve that starts out looking like Curves in darken settings, takes on a more "S" shaped look, due, I suspect, to normalization.
I wish Photoshop had a similar input/output display of TF.
George: I just couldn’t resist my first troll comment, but the level slider uses a function where curves use an algorithmn.
To make this more fun, a function is part of an algorithmn, but not vice-versa.
ID.
Semantic distinctions between ‘function’ and ‘algorithm’ are beside the point—the point being that Levels uses a discreet fixed formula (with adjustable parameters, of course), whereas Curves does not.
The Curves output obeys no preset algorithm/function/formula. It can take any shape, as manipulated by the user. The Levels output is constrained by the tool’s designer, rather than the tool’s user, who is left to tweak sliders—albeit over their full ranges.
And neither tool is restricted to "linear" action—the essence of your contention.
George
No George, I won’t tell! Just musing! I like the info as well.
As both technical person and photographer, I appreciate the way PS seems to hit the bullseye on so many things. My comfort level is much better than with casual markings. But to be honest, I am in that "set according to looks" position. I use the numbers to set certain functions like corner burndown quickly. (.9 is usual) It is more repeatable and quicker than Curves, where you have to pick the point and figure out where to go. The rest of the time, the only value to knowing the slider (gamma) is past a certain point, I know I need to take a look at Curves.
This is for global only. For major interpretive efforts, I never use Levels.
I do think a dynamic presentation of the transfer function in the Epson is even more telling. That I will recommend to Adobe.
Gee George, that was a really funny answer, I know you didn’t intend it to be, but that was side-splitting hilarious. Read it again, boy, read it again.
OK, I read it again. And again.
George,what’s so funny?
butler…..
Lawrence—You’ll have to ask ID. I’m not even sure to which message he is referring. Maybe they’re ALL funny—have I missed my calling? Up til this very recent success, I’ve been accustomed to groans… Now then, James, please place a call to my agent…
George
I would be grossly OT had not this thread veered off into the hilarious. I take license from that to call to your attention a baseball stat that is so wondrous it IS hilarious.
Michael Lewis in "Moneyball" (p 270)points out that the 2002 American League West teams finished in the standings in exactly the reverse order of their payrolls.
If you look a little closer, the payroll increments between teams is about $21M (give or take a small percentage)per rung in the final standings.
Now, the AL West is the only one of 6 MLB divisions to have only 4 teams. All others have 5 teams.
Add a mythical fifth team to the AL West, give it an annual payroll of $21M, and postulate that the standings vs payroll rule for that division in that year would still have held. Result:
1st place $ 21M payroll Fifth Team
2nd place $ 42M payroll Oakland
3rd place $ 63M payroll Anaheim
4th place $ 84M payroll Seattle
5th place $105M payroll Texas
Maybe inverting the exponent in the gamma slider box is a sign of the times!
George
