Elle Stone's tutorial, Autumn colors


(Gustavo Adolfo) #21

I think @afre either converted to 32-bit floating point and buit-in sRGB, just in a different way: he opened a blank image file in 32-bit float and then opened the original xcf file as layer. This will convert to what we want.
I guess.
Just don’t understand the reason for that workaround

(Bill Martz) #22

To apply the V4 RGB working space profile:

C1.3.2 Extract the B channel using “Colors/Component Extract” and select the LAB “B” channel. Change the layer name to “LAB B channel”.

C1.3.3 The resulting extracted layer will have the wrong tonality because the “Component Extract” procedure doesn’t take into account the TRC of the layer stack. So drag the B channel layer out as a new layer and assign to it a V4 RGB working space profile from my profile pack that has the “LAB L” companding curve. These profiles all have file names ending in “-labl.icc”.

Drag the resulting layer back to the original layer stack. Go ahead and delete the originally extracted layer.

I dragged the “LAB B channel” layer down to the bottom of the whole stack and changed the profile there. Then I dragged it back.

(Bill Martz) #23

C1.3.4. Drag the resulting layer back to the original layer stack. Go ahead and delete the originally extracted layer.

I took originally extracted layer to mean the layer labelled “scene-referred for LCH Chroma group.” At the end of step C1.3.6 my stack looked like this:


And the image looked like this:


In that image, I see a lot of bright orange, green, and some yellow pixels that are not in the Blended result Elle showed. Her image also appears to me to show effects of the Grayscale mask, but I don’t see that in my image.

My discrepancies only get worse from here once I start working on the Lightness group layers.

(Gustavo Adolfo) #24

How did you do that? In Image -> Color Management? If so, you’re not only converting that layer’s profile, but all layers. I think that’s not what Elle’s meant to do. In that case, @afre’s second workaround (Image -> Duplicate) makes sense.

(Bill Martz) #25

I never checked the other layers until now, and you are right. They are all sRGB-elle-V4-labl.icc.

I have no idea how that affects things. I will have to look at @afre 's second workaround and try it out.

I’d like to see what his image at the end of C1.3.6 looks like.

(Pat David) #26

I just heard from Elle. She’s seen this thread and had started a revision of that post a while back. She’s hoping to finish it and update it later today for GIMP 2.10 (w/o referring to GIMP CCE). The screenshots may be out of date, but the text process should reflect the updates.

(Gustavo Adolfo) #27

Well, I’m still at the very beginning of the learning curve, but changing all layers to a smaller collor gamut will certainly affect it (in a bad way).

And good news from @patdavid, above (Thanks, Pat!). You made it, @Underexposed!

(Stefan Schmitz) #28

Autumn is a state of mind - wind, grey sky … even in July

Complete post-treatment: colors desaturated.


However, as you can see from the screenshot, GIMP thinks sRGB-elle-V4-g10.icc is perceptual gamma (sRGB TRC). We don’t want to go from perceptual to linear during depth conversion. We only want depth conversion.

Screenshot again, for your convenience.


(Gustavo Adolfo) #30

@beachbum this was edited according to Elle’s workflow?

(Bill Martz) #31

That’s the best news @patdavid. Thanks.

Tried doing this, but not sure how the layer’s image was copied and put into the original image stack. The result of all the maneuvering seems only to be a selection mask on the Visible + Channel mixer to add Chroma layer, since when you delete the “LAB B channel” layer, any other effect of it disappears. I followed the original instructions, including the profile change, but once I had the selection mask, I changed the profile back and deleted the “LAB B channel.” Didn’t seem to help.

I look forward to the updated tutorial.

(Bill Martz) #32

I did learn why the result of step C1.3.6 does not match Elle’s screenshot of the blended result. In the instructions for extracting the LAB B channel, she did not include, “making sure to check the ‘Invert component" box and the "Linear output’ box,” as she does elsewhere when discussing making a Chroma mask. (https://ninedegreesbelow.com/photography/changing-saturation-using-lch-chroma.html)

The bright oranges, yellows, and greens are gone.



I am not prepared to break down the steps for you right now. I have had a tiresome week and discuss has been a good distraction. I think I have left enough cookie crumbs. Maybe someone else could lend a helping hand. :wink:

Your conclusion is wrong. The purpose of the mask is to isolate the organics so that you may tweak their chroma. If you invert the mask, then you would be working on the inorganic regions.

(Elle Stone) #35

Hi All,

My apologies, my Autumn Colors tutorial was written in 2015. Since then so much has changed in default GIMP, rendering the original tutorial no longer workable even in later versions of GIMP 2.9, much less in 2.10.

The original 2015 version of the tutorial used “Colors/Components/Decompose to LAB” to make the original Chroma mask, using the “B” channel. But when I started updating the tutorial several months ago I switched to “Colors/Components/Extract Chroma” thinking this would be easier. But actually this left the tutorial completely “broken” in the steps to increase Chroma, because it turns out that extracting the Chroma channel clips channel values to the range 0.0-1.0f.

In the last two days I’ve updated sections A, B, and C. I just now uploaded the new text and screenshots and also uploaded new downloadable xcf and curves files. I’ll work on the remaining sections over the next days.

If @Underexposed hadn’t started this thread no doubt many more months would have gone by before I noticed anything wrong with my “Autumn Colors” tutorial. Which would have been a shame because the techniques that are described really are useful for editing and also for painting, and AFAIK no other image editor than GIMP allows quite this type of editing. So @Underexposed - thank you! for having the patience to work through the tutorial and explain the problems.

Hopefully the revision is a lot easier to follow, and if not please let me know and I’ll attempt to remedy the remaining issues. Various people in this thread have raised various issues, some of which I’ve tried to work into the tutorial, and others I’ll try to answer here in this thread.


(Elle Stone) #36

Hi @afre and @Underexposed - actually you are both right!

In the original 2015 tutorial I used the LAB “B” channel as the initial Chroma mask, to isolate the sky, as the sky was very positive in the “B” channel and somewhat in the shadows cast by the trees, but not elsewhere. The “B” channel mask made it easy to target the amount of saturation added to the sky, by adjusting channel values in the B channel that were greater than middle gray. And then I used the Chroma channel to make selections to further modify the “B” channel mask to lower the highest Chroma colors in the orange and yellow leaves on the trees. This was a fairly complicated way to make a Chroma mask, and I didn’t explain all the steps even in the original tutorial. I just provided the final Chroma mask instead.

In the revised tutorial I skipped using the “B” channel entirely and just used the LCH “C” channel to make the initial Chroma mask, which requires inverting the “C” channel to mask the highest Chroma areas more than the lower Chroma areas.


Definitely, it would depend on what you are doing and what your goals are. Now that you are back, I will retire from customer support. :sunny: Let us know when your updates are complete.

(Andrew) #38

@Elle, good to see the yellowy-green E circle again (even if just a brief visit)!

(Bill Martz) #39

I have been buried with work and family activities, and I haven’t had time to look at the revision. But I thank Elle for it, and will try it out as soon as possible.

(Bill Martz) #40

Having worked through the new tutorial several times, documenting my work to catch mistakes and verify when I got it right, there’s a couple items I thought I would bring up for discussion. Before I do that, when I completed the tutorial, I was very pleased with the results—they matched, as far as my untrained eye could see, the after images in the tutorial. Thank you very much, Elle. I’ve really learned a lot.

First, just to mention that on step B.3.4, Figure 7. indicates that Auto Stretch Contrast (ASC) is used, but the text does not mention it. I tried it both ways and compared the histogram against the one in Figure 8. There’s a fair amount of difference between the two histograms of with and without ASC, and it seems to me that the one when ASC was used matches the histogram in Figure 8.

In step C.2.3, it says, “A layer mask is used to avoid blowing out the brightest portions of the sky.” The layer mask, Mask for Curves for sky, includes a selection. The text does not indicate that the selection be applied to the image, but I assumed that it should be because an important goal was to prevent blowing out the lower left portion of the sky. Therefore, sample points 2 and 3 should not, I think, change when the curve is applied because they are outside the selection area. However, when I completed this step the RGB readings for all three sample points changed.

C.2.3 Mask for Curves for sky


C.2.3 Before:


C.2.3 Curve application:


C.2.3 After:


All three sample points changed.

Similarly with step C.2.4, the sky is to be protected from the curve application:

C.2.4 Mask for Curves for ground (from Blue Channel)


C2.4. After curves application



None of the sample points should have changed, but 1 and 3 did.

In both cases, I tested the selection usage by creating my own selection area that was roughly the same as the tutorial’s, and in both cases, the sample points behaved as expected.


I’ve already shown two of the masks used, not counting the Chroma mask. I wanted to figure out how they were made, as they are extremely important to managing clipping. I tried out the method given by Pat David in a tutorial on Luminosity Masks, using luminance instead. What I was especially interested in determining was how to get the tonal variations in the sky. However, none of the luminance masks I produced revealed the variation found in Elle’s mask.

The mask for ground cover indicated in its name that it was from the blue channel. Below in the blue channel I pulled from the image after step C.2.3. I am puzzled as to how the mask shown above was made from that.


This is C.2.5 Mask for High pass. Just guessing: came from green channel, used curves to darken and then selected the dark areas, saved the selection, used curves to get the tones in the trees, applied the selection, filled with black, sharpened in some way. ???


Referring back to the sky mask, I can get close to getting the white areas, but the tone and, in particular, the tonal variation in the grey area, how was that done and how was the variation decided upon?


This is just a quick reply. I have a feeling that the selection is a soft one meaning that the sky should change slightly. Otherwise you would have harsh transitions. However, it does look like the change is a bit too much and it isn’t evenly distributed among the colour channels, which is an issue…