Pages: 1 [2] 3   Go Down

Author Topic: levels  (Read 14009 times)

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #20 on: June 15, 2005, 10:30:17 pm »

Jonathan, Jack et. al.

OK, I have gone to the Temple Mount, communed with an authoritative Guru (who will remain un-named because I didn't ask permission to attribute) and thought further about this myself.

Bottom line: in principle you are correct: any curve shift alters the color balance, BUT in practice the impact depends on the specific situation. Quote: "If you move the endpoints of the individual channel curves in such a way that the midpoints don't move much, it won't affect midtone color balance." This is probably what was happening in the tests I performed, because on the other hand - quote: "Anytime you make differential edits to individual channels, you'll get color shifts somewhere. That's usually the reason for doing so, in fact."

It was an interesting exercise and somewhat insightful.

Cheers,

Mark
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #21 on: June 16, 2005, 09:00:41 am »

Thanks for the acknowledgement Jonathan - no harm attaining a bit of wisdom now and then - and I think the points you added also answer Peter McLennan's query with the added clarification that unless the shapes of the histograms of the channels underlying the composite are identical, when you tweak the composite channel that tweak is a uniform move operating on each of the underlying channels that by definition have different starting structures, hence the actual resulting change to each channel will differ thereby altering the color balance.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

ByronWill

  • Newbie
  • *
  • Offline Offline
  • Posts: 1
levels
« Reply #22 on: July 13, 2005, 01:28:48 pm »

Hi,
Anyone care to explain the difference in terminology between tonal range and dynamic range? I assume that dynamic range relates to bit depth from black to white, where tonal range refers to just the range from black to white?? These terms seem to be used interchangably at times.

Thanks, Byron
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #23 on: June 12, 2005, 03:09:53 pm »

It isn't feasible to show 65000 levels in the chart, but if the image identification pane at the top of the Photoshop window says RGB16 you are nonetheless working in "16 bit", but in fact Photoshop in its 16-bit mode works with 32000+ levels, not 65000+. I don't know whether or not the conversion process from 12 bit images to 16 bit mode leaves or smooths gaps, but you wouldn't see them anyhow, the gradations are so many. In levels you adjust the composite RGB channel for changing overall luminosity (contrast and brightness). You would adjust the individual R,G,B channels for two reaons: (1) also to change luminosity, but making sure you don't clip any highlights or shadows in any of the three channels, or (2) to intentionally rebalance colors of an off-color image. "Curves" gives you much more control and flexibility for all of these purposes, but using Levels is often sufficient and convenient.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
levels
« Reply #24 on: June 14, 2005, 11:36:25 am »

Quote
I thought that adjusting the RGB's indivually would be good for changing the luminosity, however, when I tried moving the levels of each colour individually to their upper and lower limits (without clipping any of them) it alters the colour balance. Do you have any idea why this would happen?
That would be because altering the value of one color channel vs another is how you get different colors in a digital image. Your question is sort of like asking why fire is hot.
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
levels
« Reply #25 on: June 14, 2005, 07:40:25 pm »

Quote
Think back when you first got into digital photography - was all this stuff so immediately obvious to you, or did you have to read, learn, think and experiment?
In this instance, yes it was. Given that just about any color perceivable by human vision can be reproduced by altering the mixture of red, green, and blue color channels, it is totally illogical to think that altering one color channel by itself would not affect the color balance of an image. It's fairly simple logic:

Premise 1: RGB images encode a wide range of colors by mixing various amounts of red, green and blue. You can change one color to another by changing the red, green, or blue color values.

Premise 2: Using the levels control on one color channel in an RGB image will change the color values for that channel compared to the other channels.

Conclusion: The levels control, when applied to a single color channel, will change the color balance of the image, because the values for one color channel are being increased or decreased in relation to the other two color channels.

Think back to the recent thread about polarized sunglasses. The original poster claimed to be familiar with the operation and effects of a polarizer, and said he knew that his sunglasses were polarized, yet was sufficiently amazed by the fact that the polarized lenses actually did what they were designed to do that he felt it worthwhile to post his "discovery" to share it with the rest of us. But again, simple logic should make that instinctively obvious:

Premise 1: Polarizers can reduce or eliminate reflections on water.

Premise 2: My sunglasses' lenses are polarizers.

Conclusion: My sunglasses' lenses will reduce or eliminate my ability to see reflections on water when I wear them.

I don't mind answering questions that indicate the questioner put some thought and effort into posing the question, but when questions are asked that indicate either a total lack of understanding of the most basic, foundational concepts, or (more often) a near-complete non-use of critical thinking skills, it makes me (and others) feel like the driving instructor whose student asked how wheels work. At some point, it becomes difficult to distinguish legitimate questions from silliness posted with the sole purpose of being annoying. I know people say there is no such thing as a stupid question, but that isn't true; there really are stupid questions:

"What's the number for 911?"
"Do you think I should let my kids sleep with Michael Jackson?"
"My stripper girlfriend is cheating on me. If I married her, do you think she'd stop?"

I'm sure nobody would find it particularly helpful or useful or relevant if I posted a thread about how the sun rises in the east or how rocks are a bad choice of lens cleaning implements; where do we draw the line? I don't hate anyone, the primary goal of what I write is to get people to think logically and use more common sense. A little bit of that would go a long way toward answering a lot of the questions posted here and elsewhere.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #26 on: June 14, 2005, 11:39:42 pm »

Yes, OK, he could have done some more homework, and Yes,  many people would find it obvious that recomposing color channels changes pixel values, while some don't; BUT THAT IS NOT THE POINT. No need to move on guys - just CHILL-OUT and be cool.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

paulbk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 528
levels
« Reply #27 on: June 15, 2005, 12:43:45 am »

hovis,
It takes some understanding of what the Levels histogram is telling you in order to anticipate the likely result of a manipulation. Let’s look at what happens to a grey pixel when we fiddle with an individual color channel in the Levels dialog. Assume our example pixel is 8 bit RGB mid-grey, that’s R=128, G=128, B=128. Now select the red channel and move its mid-point to the left about half way to black. Photoshop will redistribute all pixels with a non-zero red value to a new brighter/higher red value. So our grey pixel now looks more red since the old value R=128 has been reassigned a some new higher value (~R=180 or something close to this). Also note that the Green and Blue values have not changed.

Try it. Pick any file you like and locate the Histogram pallet so you can see it while you manipulate the Levels dialog. Use the Histogram flyout menu and turn on [All Channels View] and [Show Channels in Color]. Now open the Levels dialog, select the Red channel and move the mid point up and down the scale. Watch what happens to the red channel histogram. Note this changes the white balance of the whole image since you are changing the relationship of RGB for each pixel.

I call it learn by fiddling. After four or five-hundred fiddles even I get it.
paul
Logged
paul b.k.
New England, USA

Eric Myrvaagnes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 22814
  • http://myrvaagnes.com
    • http://myrvaagnes.com
levels
« Reply #28 on: June 15, 2005, 01:13:16 am »

Quote
In case you didn't recognise yourself Eric, the first paragraph was an impersonation of you. You can take comfort in the fact that I've sunk to the right level.
Dear Hovis,

I am sorry if my post offended you. Perhaps I was too subtle. I was not trying to insult you, I was simply acknowledging that Jonathan sometimes seems to have a short fuse, which is really too bad, because most of his posts are truly helpful.

The fact that you felt so insulted by my post really astonished me, so I went back to reread the entire thread to see what had happened. As I see it, you asked two very reasonable questions in your first post, and MarkDS gave what I thought were good answers to both. You then thanked him, and added a new question that did suggest to some of us that you really needed some basic help with color theory. MarkDS again gave a thoughtful response, as did Bernard.

Then Jonathan made his comment, that seems to me to be a bit over the edge. But your reply to him sounds to me as if you are escalating the conflict. Your comment and MarkDS's next comment both sound to me less courteous than Jonathan's original comment; nevertheless, he graciously apologized.

At this point mcanyes jumped in with an even more insulting and, to my way of thinking, unnecessary comment. The whole thread (at least the name-calling) should have stopped there, but it didn't. Jonathan tried to explain where he was coming from, and in so doing he poured a bit more gasoline on the fire. I added my comment, which was milder than any of the previous six comments (except for Jonathan's apology -- note that he is the only one who has apologized so far, except for me, and I really didn't mean to belittle your requests for information, so again I apologize.)

Jack then put his two cents in in support of Jonathan, who was being rudely pilloried by this time. And MarkDS rightly suggested it was time for all to "just CHILL-OUT and be cool."

You then took two more posts to insult me, when I was in no way trying to offend you.

One of the nice things about this forum is that there is, indeed, a place for questions at all levels. Stef-T is a beginner who has asked many good, beginner-level questions in a way that have prompted helpful responses. And there are certainly times when I have come close to asking questions that will bring down Jonathan's scorn upon my head. But I strongly urge you not to take any of this personally. For some people (such as Jonathan), a lot of these technical things do come really easy, so they have a hard time understanding why it's not so easy for other people. My brother is very much like that, and he can be a real pain in the sensor. In Jonathan's case, his posts that have good, accurate, useful information greatly outnumber the ones in which he loses his cool. I strongly recommend simply ignoring the latter and paying attention when he does say something you may find helpful.

And I hope you can lift yourself above the level you felt it necessary to sink to in responding to my post. And good luck getting a handle on the color management stuff. I for one don't find it easy.

Eric
Logged
-Eric Myrvaagnes (visit my website: http://myrvaagnes.com)

jani

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1624
    • Øyet
levels
« Reply #29 on: June 15, 2005, 08:13:16 am »

Quote
(...) and just joined the site this month, I'd say the manners should start with you.
Just a minor comment:

That someone registered at a particular date doesn't mean that one hasn't been reading for a longer period.

I think it should be allowed to be a "lurker" and then register when you desire to post something, and not just because that's what I did. It's a long-standing tradition on most bulletin boards, going back a couple of decades.

Also, how long one's been posting is no excuse for poor manners.

Poor manners are poor manners, no matter who's displaying them.
Logged
Jan

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #30 on: June 15, 2005, 12:04:32 pm »

I would like to suggest to my virtual colleagues that we put the subject of civility to rest - as it has been exhausted - and at the risk of getting Jonathan and Eric's dander up we return to the technical aspects of Hovis' basic question, because it isn't either that dumb or that obvious. To quote Hovis from near the start of all this:

<<I thought that adjusting the RGB's indivually would be good for changing the luminosity, however, when I tried moving the levels of each colour individually to their upper and lower limits (without clipping any of them) it alters the colour balance. Do you have any idea why this would happen? I'm sure I'm missing something but it seems to me that unless one actually encroaches into a colour (thus removing active pixels) it wouldn't be possible to alter the colour balance of the whole image.>>

You see, he wasn't asking about shifting mid-points, he was asking about the impact on color balance of filling the luminosity range by stretching each channel to the end-points without clipping, and wondering why when he did this the colour balance gets altered. None of us have done Hovis a service in the technical answers we've provided, me included. The fact is that his basic premise is correct - doing what he did should NOT alter the colour balance of the image, only the luminosity. I think what happened to Hovis is one of two things: (1) when he saw his image fall into a proper range of contrast and brightness he mistook this for a change in colours and/or (2) his adjustments to the end-points of the histogram may have looked accurate on the monitor image of the histogram but may have been slightly off in the actual numbers.

To satisfy ourselves that what I am saying here is correct, the empirical test is best implemented in Curves, where we have the detailed data we need for this and we can replicate the identical moves we make in Levels by shifting the white point to the left and the black point to the right (assuming black is lower left) for each channel up to the point that there is no clipping of anything anywhere. I selected a RAW image that failed to fill the luminosity range (especially all values much too far below the white point) and has enough grey in it to easily anchor a colour-neutral point. So I dropped an anchor on a natural grey point that had a value of level 171 in each of the RGB channels - perfect. The hypothesis is that if exactly filling the luminosity range of each channel without clipping alters colour balance along the range, then my resulting neutral grey point should change to unequal values of R,G,B. To test it, I very carefully adjusted the white and black end points of each channel's curve as far as I could without clipping; then I went back to measure the value of my anchor point: 235, 235, 235 bang-on neutral grey but brighter.

Then I selected another more colourful image and replicated the same procedure selecting a coloured anchor point and observing the impact on percentage composition of each of R,G and B to total RGB for that colour. My pre and post-adjustment percentages compared very closely, with slight variances I suspect due to limitations of visual accuracy in adjusting the end-points to the clipping point.

The discussions in Blatner/Fraser and Eismann referenced elsewhere in this thread don't really address this issue directly. Blatner/Fraser (page 306 onward) focus on what happens when you make adjustments to the COMPOSITE curve where clipping does occur in the underlying colour channels, while Eismann mainly discusses using Levels for color correction.

So Hovis, your initial premise seems correct; therefore, you may wish to consider the other reasons I suggested about why perhaps you didn't see what you thought you saw!
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
levels
« Reply #31 on: June 15, 2005, 02:42:53 pm »

Quote
Shifting midpoint via the center gamma box or changing the black and/or white point doesn't matter. If you do any of these things to one color channel, you are altering the values of that channel relative to the other channels and will get a color balance shift.
Yes of course you are correct in this useful clarification, but just to remind - that was neither the question nor the full content of the test. It is about what happens when you adjust all three channels so the end-points of the three histograms are just at the clipping frontier.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

sc21

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 74
levels
« Reply #32 on: June 15, 2005, 09:09:41 pm »

Hey hovis, if you're still reading this, you wrote:

I was wondering why; when working in 16 bit mode in both C1 pro and Photoshop are there only 255 tonal increments shown in the levels chart instead of the 65,000 or so that exist.

I just read a bit on this the other day in "Real World CS":

"For outrageous precision, CS even displays the 16-bit values - ranging from 0 to 32,768 - when you work on high-bit files."  It also mentions how they still work with the 8-bit values "at least until we get used to thinking of midtone grey as 16,384."

So there's gotta be a way to turn that on.  Just checked, and while it seems you can't get it in Levels (the boxes aren't big enough to show those numbers), you can go to the Info Palette - click on the little arrow for Options - and select "Show 16 values."
Logged

hovis

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 84
levels
« Reply #33 on: June 16, 2005, 09:03:26 am »

Thanks for all your replies, especially Mark for taking the time to really look at this.

It's been very interesting for me and I've learnt a thing or two, Jack's suggestion with the curves tool is very useful.

Thanks again all.
Logged

hovis

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 84
levels
« Reply #34 on: June 12, 2005, 03:36:55 pm »

Thanks for your reply.

I thought that adjusting the RGB's indivually would be good for changing the luminosity, however, when I tried moving the levels of each colour individually to their upper and lower limits (without clipping any of them) it alters the colour balance. Do you have any idea why this would happen? I'm sure I'm missing something but it seems to me that unless one actually encroaches into a colour (thus removing active pixels) it wouldn't be possible to alter the colour balance of the whole image.
Logged

hovis

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 84
levels
« Reply #35 on: June 14, 2005, 01:52:43 pm »

Hi Jonathan

learn some manners, if that's the best reply you can come up with don't bother. If you'd venture into someone elses world maybe you'd realise that you're not as bright as you think, or have you already tried that and this is the reason for your lack of courtesy. Things aren't always so straightforward for the uninitiated and remember that it's a person not a computer you are talking to.
Logged

mcanyes

  • Full Member
  • ***
  • Offline Offline
  • Posts: 119
    • www.dig-arts.biz
levels
« Reply #36 on: June 14, 2005, 04:02:00 pm »

Jonathan,
You are the poster child for bad behaviour. There are a lot of comments on this forum, and other forums that are unnecessary.
Michael
Logged
Michael Canyes
Nikon stuff www.dig-arts.

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
levels
« Reply #37 on: June 14, 2005, 11:25:55 pm »

Hey guys...

I have to chime in on this one and defend Jonathan on this particular question.  Simply stated, the question shows that the poster obvisouly did not even make an attempt to uncover a basic answer before posting his question.  

Had he simply clicked the "help" button at the top right of the tool bar in CS, and then navigated to "index" then to "curves command" he would have seen several subheadings that give a great deal of insight to what the tool can do.  Then had he clicked on one of the subheadings like "adjusting color balance and tonal range with"  -- and gee, do I need to point out that this was more or less his original question -- he would have seen this as an initial explanation:

"The Curves dialog box, like the Levels dialog box, lets you adjust the entire tonal range of an image. But unlike Levels, which has only three adjustments (white point, black point, gamma), Curves lets you adjust up to 14 different points throughout an image’s tonal range (from shadows to highlights). You can also use Curves to make precise adjustments to individual color channels in an image. You can save settings made in the Curves dialog box for use in another image. See To save and reapply settings in a dialog box."

Maybe Jonathan and I need to move on...

Cheers,
Logged
Jack
[url=http://forum.getdpi.com/forum/

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
levels
« Reply #38 on: June 15, 2005, 12:33:37 am »

My ultimate goal here isn't so much to be the resident curmudgeon as to encourage people to learn how to fish instead of simply giving a fish to everyone who requests one. If people wish to debate my choice of techniques, that's fine, but I think we can all agree that most internet forums would improve significantly if more people spent a few moments checking the applicable help file or using the forum search tool or simply thinking through their question to see if it makes logical sense before posting.
Logged

Eric Myrvaagnes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 22814
  • http://myrvaagnes.com
    • http://myrvaagnes.com
levels
« Reply #39 on: June 15, 2005, 01:20:10 am »

Paul,

That's a great experiment. I like your "learn by fiddling".

Eric
Logged
-Eric Myrvaagnes (visit my website: http://myrvaagnes.com)
Pages: 1 [2] 3   Go Up