My eyes are failing or is the end nigh

[leeharper_admin]

I reckon we have flogged this poor rose to death!

Oh, I wouldn't be so quick to say that ;)

Your screenshot is very interesting Chris. I hope that you will be able to answer a few questions that I have about it:

• Your Munsell step wedge is divided into 11 segments, whereas the Lab version has only been divided into 9 segments. Why is there this disparity?


• Your Munsell values don't appear to follow the shape of the 5R colour swatch. It looks as though you have derived values between 0/0.7 and 10/7.3 along a straight line. The Munsell swatches show that chroma (saturation) should be highest at value 6; in your step wedge it's highest at value 10, and then it gets progressively lower. Why is that? I'm wondering if the Lab step wedge looks more realistic because the Munsell step wedge is not respecting the shape of the 5R colour swatches; am I missing something?

Your zone step wedges look really promising. I would like to understand them a little better...

All best,
Lee.

[mikemeister_admin]

Hi Lee,

1. The Munsell Saturation ruler is in 11 equal segments in steps of 10 L (0 to 100).  The Zone System Ruler follows the system defined by Norman Koren (http://www.normankoren.com/zonesystem.html) which I believe is more in line with the idea of the Adam's Zone system, than equal segments, for digital images and has less segments than the original 0 - X.

2. The advice by Frank Reilly to painters on how to mix colours as the object became darker is generally accepted as correct and follows the old masters. Basically, as an object colour in a scene being painted becomes progressively darker, its Munsell value, not surprisingly, decreases. The rule of thumb says that the chroma also decreases, by about the same amount proportionally.

Although it seems counterintuitive, the hue of a shadow is not nearly as important as its chroma. As the shadow becomes darker, and the chroma decreases, the hue becomes nearly indistinguishable.

Paul Centore has written an article (http://www.99main.com/~centore/ShadowColoursForPaintersLowRes.pdf) that explains all this, but in a nut shell it says...
Suppose that the colour when lit has Munsell coordinates H V/C

For example 10GY 7/12.  The left axis is a vertical line of neutral greys, numbered from N1 through N9. Continue the neutral axis downward, to where the value would be -1.
Draw a line from the value -1 on the neutral axis, to the center of the square containing 10GY 7/12.
The shadow colours of 10GY 7/12 fall on this line.

This is obviously approximate as the chances of a colour exactly matching a Munsell swatch is not that high.  The Ruler swatch values use this formula, starting with the selected colour (as denoted by the '*' swatch).

I find it quite informative to use these rulers on images and especially to guess/check Zones when dealing with coloured objects - I find it hard enough with grey-scales, but colour is a whole different ball game!

My odd experiments seem to agree that the Reilly saturation model creates 'better' looking images than straight hue values, especially when we boost colour (for impact effect) compared to the original.  Without a doubt simply using a L mask during Lab colour saturation helps our perception.

[leeharper_admin]

Hi Chris,

Thank you for your reply, your comments are very helpful.

Frank Reilly's advice seems to conflict with the Munsell system; either that or I'm getting in a muddle. I've attached an image that I found online. It demonstrates that chroma should not decrease in relation to value for all colours. As we can see, if we are dealing with yellow, chroma decreases with value, but if we look at blue (10PB), chroma increases as value decreases, and for red chroma first increases, and then decreases.

Am I missing something?

Lee.

[mikemeister_admin]

Have a look at page 4 for Paul Centore's article and hopefully that will explain it

The images you picked up is not correct and show a lot of out-of-gamut colours

I think the attached (in gamut only colours - on 2 backgrounds) are more like it - but I think you are not using the Munsell terms correctly (http://en.wikipedia.org/wiki/Munsell_color_system)

[ggroess]

Yes Greg a much nicer tonal range, but now the rich inner colour is lost.

To me this is one of those cases where getting the B&W tones correct does improve the image.  If I put "gg1" over the original in luminosity blend mode, the colours are still lost at the highlight end and even using a L channel mask during Saturation boost does not completely help because of loosing the very fine detail in the light petals.

Interestingly on my system the center color is a bit over the top.  The posted image looks less saturated but I'm sure the file is correct.  I was worried that I had over stated the center of the flower and lost the inner details.  Mostly I wanted the green gold on the leaf stems and the pink hue to be consistent across the outer petals of the flower without losing the richness. 

re: the end is nigh...

I'm thinking that as cameras get better our job will change but not necessarily go away.  If you relate back to analog film it came down to process controls and exposure.  Then there was the custom printing services that made the image.  I do not see these things going as far away as we once thought.  I'm thinking that the more things change the more they are staying the same in this regard.  We are having to wrestle with smaller and smaller technical differences as the image capture gets better. (closer to film response or even something as a new standard)

Once we all "went digital" we adapted to the new processes to try to make images look as good or better than we did before.  When we hit the outer limits of the medium we were using we accepted the failure as out of our control or needing a better algorithm or technique to solve the problem.  We even felt at times we  needed better hardware or the latest thing (marketing heaven).  Now I think we are circling back to the original issues with Photography; here it is becoming more about subtle controls that effect the realism of the image and perception.  The translation of captured image to the analog print in some cases is more difficult because we use our home printers and everyone is on their own for controls. 

I see our job in the future as possibly being one of a more nuanced response process, what I mean here is we will be responding more to subtle issues and less to major disasters image wise.  Dan M teaches people how to "save" a really damaged image.  I think we are getting past that technical issue as cameras get better and we are now dealing with a finer polish process rather than the "bigger hammer" to drive smaller nails.

2 pence please... 8)

Greg

[leeharper_admin]

Hi Chris,

I have read through Paul Centore's article - thank you for the link. Centore's rule of thumb then is that the chroma of an object decreases in a linear fashion from the local colour of the object in question. That's reasonable, and his examples are convincing.

Centore's article does not deal with how chroma should be treated in areas brighter than the local colour of the object. Presumably the problem of overly saturated shadows would also occur in highlights if the chroma wasn't modified. I notice that Paul Centore's article carefully states that the object he is discussing has a matte rather than glossy surface, and that it is being lit indirectly. In those circumstances everything's fine; however, the world rarely plays to those rules, so to really make use of the Munsell system we need to know not only how to modify colour in areas darker than an object's diffused value, but also what to do about brighter areas.

How does your Munsell zone ruler deal with this? I notice on your rose screenshot that your ruler defines three zones brighter than the value you are reading. Does Frank Reilly discuss this problem?

Thanks for your advice,
Lee.

[mikemeister_admin]

Hi Lee, yes you seem to have noticed everything and as you say it is all approximate.

However the idea is there, but the precision is not.  I personally do not believe that we can accurate match an RGB value to Munsell as it is down to interpolation and many other issues.  I therefore reckon that all one can do is go for the closest swatch and then work from there.  If we assume that the Reilly formula is correct, then the line must logically work for brighter tones as well as shadows!  (We could have picked a brighter instance in the first place and then seen that it falls on the darker one down the Reilly line). 

If I remember correctly, the ruler also include out-of-gamut colours, so it is not accurate. 

The more I look at sRGB colours the more surprised I am about how limited they are, but then I know from trying to reproduce an image of a painting that it is virtually impossible to get the correct colour - so all my post-processing is making the colours reasonable (to my screen/print & eye) rather than accurately reflect the original object (and lighting conditions etc).

BTW, don't thank me for my advice - I am a bumbling amateur who likes to explore things and then attempts to draw conclusions from what I find/read.  I sometimes think that the conclusions may be wrong, especially with regard to our eye/brain combination.  I am (rapidly) giving up on post-processing by numbers and using my eyes instead as the Hue clock!  This is a dangerous road to travel, especially if the monitor is not well calibrated and the eyes have 'adapted' to tints.  In terms of screen images (which is what we mostly talk about) the difference between CRT and LCD renditions is quite marked - so I look at both and try to navigate between them.  I still use the Hue clocks - would never be without them - but I (frequently?) move away from setting a digital neutral in favour of a perceived neutral for the colours and objects around it.

So as you can guess my photos are beginning to lie more and more - and hopefully most of the time nobody notices.  Hell - photography is an art form after all.
Chris

[ggroess]

You have to remember Chris that the choice for a neutral unless it is 100% known is always subjective.  We should be choosing a point that we believe to be neutral.  From that point forward we have bet the image on it.  I find my color to be better overall if I start with a highlight I believe to be neutral and a shadow I can "Accept" to be neutral. 

Highlights tend to show a color cast to me better than Shadows.  My brain accepts the shadow cast as normal even when I know it is not; I have to always check them.  I always measure but have become less focused on exact numbers.  I'm relaxing about that one except when I am teaching CM101.

Better to have a known starting point in that arena.

Greg

[leeharper_admin]

Morning Chris,

If we assume that the Reilly formula is correct, then the line must logically work for brighter tones as well as shadows!  (We could have picked a brighter instance in the first place and then seen that it falls on the darker one down the Reilly line).

Calculating the saturation level (chroma) of brighter tones by continuing along the Centore line would result in saturation increasing in areas brighter than the local colour of the object in question - whereas saturation actually decreases (see the attached images of the flags - borrowed from: http://www.lloydgodman.net/tech/tech/Exposure/saturation1.html, and also the attached image of the bananas - borrowed from: http://www.wetcanvas.com/forums/showthread.php?t=928132&page=2).

This decrease in saturation suggests that if Centore's line is to be respected for tones brighter than an object's local value, it needs to be flipped - giving us a Centore/Reilly boomerang (see my doctored version of one of Paul Centore's images).

BTW, I found a PDF which gives Lab numbers for Munsell values, that I thought you might be interested in reading. Its available here: http://docs-hoffmann.de/munsell15052009.pdf from G. Hoffman's website (http://docs-hoffmann.de/howww41a.html).

Cheers,
Lee.

[mikemeister_admin]

Thanks Lee,

I'll have to do some real experiments and get back to you, but I'm a bit tied up at the moment so it may be a few days.
I just feel that logically straight line formulas should not suddenly change direction!
BTW I believe that Yellow and Blue are bad colours to pick because of out-of-gamut problems.  Remember as Yellow gets brighter there is more Blue in it (RGB) which is clearly wrong!
Chris