hpr-knowledge-base/hpr_transcripts/hpr3520.txt

Episode: 3520
Title: HPR3520: Inversion Layer Modes
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr3520/hpr3520.mp3
Transcribed: 2025-10-25 00:56:15

---

This is Hacker Public Radio Episode 3,524 Friday, the 28th of January 2022.
Tid's show is entitled, Inversion Layer Modes and is part of the series Gimp It is hosted by Ouka
and is about 18 minutes long and carries a clean flag. The summary is, more on Layer Modes in Gimp,
with the Inversion Modes. This episode of HPR is brought to you by archive.org.
Support universal access to all knowledge by heading over to archive.org forward slash donate.
Hello, this is Ouka, welcoming you to Hacker Public Radio and another exciting episode in our Gimp series.
Continuing our march through all of the Layer Modes, and we're going to look at the Inversion Layer Modes today.
And as the name kind of implies, you know, Inversion is about inverting, okay, and in digital color,
what that means is that dark things tend to get brighter, bright things tend to get darker.
So remember, in contrast, the bright things get brighter, the dark things get darker.
Now we're going the other way, dark things get lighter, like things get darker,
but also there can be coloring versions that come out of this as well.
And again, I want to emphasize that all of these calculations that go on, go on channel by channel.
And you know, that makes a difference.
So the first one to look at is called Difference Mode.
This one's pretty simple. In this mode, the value of the top layer is subtracted from the value
of the bottom layer. And then the absolute value of this difference is used.
So absolute value means it is literally just the difference between the values
and thus it is commutative. In other words, the order of the layers does not matter.
Now, as an example, I put the red layer on top and the blue layer on the bottom
and got the magenta result when I combine them using Difference.
And if that sounds familiar to you, we got that result previously adding things together.
So why does doing the difference? Well, the math makes it clear.
For the red channel, you subtract FF from 0, 0 and get minus FF.
But then you take the absolute value and it's back to FF.
For the green channel, it's 0, 0 minus 0, 0 equals 0, 0.
And for the blue channel, it is FF minus 0, 0 equals FF.
So put it all together and you have FF 0, 0, FF.
And as we have seen, that is magenta.
Now, if we combined black and white, the result would end up being pure white for a similar reason.
The difference between them is always going to be FF, FF, FF, FF.
So in general, the effect tends to be an inversion of the colors. So I did one taking our dog image
and doing the difference with a white layer.
And the darkest part of the dog image becomes much brighter.
The brightest part of the dog image becomes much darker.
The color tends to migrate a little bit as well.
So what were predominantly brown tones in the original image are now predominantly blue tones.
And overall, it's kind of like a funny sort of a negative.
So, how does this work?
Well, let's say we have a pixel with a value of 57, 57, 8098, 57, 8098.
Well, that's kind of a dark, bluish gray.
But you combine that with the white, FF, FF, FF.
And the inversion gives us C6 AF 90.
And that's a tan color.
Well, you know, the dog had a lot of tan and brown tones in it.
So inverting, we're getting a lot of blue stuff.
So you start to see how this works.
Now, if we had used a pure black layer instead of pure white,
the dog image would be unchanged.
Now, what happens if we combine that with the toy image?
You know, a little bit of both.
The toy image had a lot of white in it.
So those places where it was white,
the dog image showing through is similar to when we just did it with the pure white layer.
Where the toy image has a dark background, it's a little bit different.
So, you know, there was a point in the both the dog image and the toy image.
There was a place on the right that was white in both images.
Now, it's black.
Now, the next note I want to look at is called exclusion.
Now, this layer mode is similar to difference.
It's just a little milder.
So I did an example.
And in my example, I combined again the dog and the toy images.
They're similar, but the exclusion mode is a little less intense.
You know, I can see, for instance, that the toy parts of the image are kind of washed out
a little more than they are with the difference.
And that's where it's most noticeable.
Well, you know, there were some dark contrasts and things there that probably added to that.
Now, this seems like a good time to bring up something that you always want to keep in mind.
And that is that we're looking at these layer modes kind of in isolation,
so that we get a sense of what the effects are.
But in practice, you would be combining different settings to get the effect you want.
So I did an example where I used the exclusion layer mode
with the toy image on top, but then I reduced the opacity of the top image to 50%.
And, you know, that's a different kind of an image.
So I bring this up because what you're going to find proficient users of GIMP
are always going to be combining a bunch of tools and settings in order to get the desired effect.
And so you just want to keep that in mind, I think, as we go through this.
Not occasionally, think to remind you of it.
Now, the next inversion mode is called subtract.
Now, this is kind of the opposite of the addition mode that we saw when we looked at the various
lightened modes. You subtract the values of the top layer from the bottom layer,
but any negative numbers that result are set to zero,
since that as low as color values can go.
The result will be a darker picture, of course, with a lot of black.
And the order of the layers does matter here.
So which one is on top makes for a difference.
So I first did one with the dog image on top.
And, you know, what I got was basically a very faint image of the dog and a very dark image of
the toy. So the dog image on top, that's the darker image, which means lower numbers.
We're subtracting lower numbers.
And, you know, you have to keep that in mind.
What if we turn it around? We put the toy image on top. Well, the toy image is lighter.
So we're subtracting larger numbers. And, you know, a lot of that toy image is white.
And so the areas where it was white or very bright are black or near black now when we combine them.
So that's how that one works.
The next one in the inversion is grain extract.
Now, similar kind of related to subtract, you start by subtracting the top from the bottom,
but then you add 128 to the values.
Now, they put that formula in the GIMP documentation, but it's obviously left something out.
You could still have negative numbers. You know, a simple example will show how that can happen
if the bottom layer had a value of say
five, a very, very dark pixel. The top had a 250.
We have a very, very light pixel. Well, 5 minus 250 is negative 245
and 128. You're still going to get a negative number.
Negative 122, I think.
So you've got to also put in a proviso that says in the event there's a negative number,
it gets just changed to zero.
Now, as it called grain, if you look at the GIMP documentation, it says grain extract mode is
supposed to extract the film grain from a layer to produce a new layer that is pure grain,
but it can also be useful for giving images and embossed appearance.
Well, I'm not sure if that explanation is terribly helpful, but you know, the best way to see
what's going on is to do some test images. And that's of course what I've been doing as I go along
with this. So I did a take first using the dog on the top image and the toy on the bottom.
You don't see a lot of the dog, but what you do see it tends to be bluish.
There's still a large area where the dog was dark that is white in this image and then an area
where on the right there was white in both of the images is now kind of a dark gray, I would say.
Now, let's flip it. Let's put the toy image on top. Now we get something that's really dark.
But you get some interesting changes. So the wizard's staff, which was kind of a darkish brown,
is now a light blue. The wizard's gloves that were yellow are now blue and so on. It's an interesting
effect. Then the next mode we can look at is something called grain merge in some sense,
the opposite to grain extract. So in this case, you add the top and bottom layers and then subtract
128 to get the final result. Now, as with the grain extract, the documentation leaves out
something. This formula is not complete because the most you can have in the result is 255.
So two very light pixels added together, manned up being pure white.
Since this particular one begins with addition, this is commutative, meaning the order of the
layers does not matter. Either picture could be the top layer and the result would be the same.
So I did a grain merge and again with our dog and the toy and you basically, you're seeing a little
of each in the image. I think the dog comes through as a little lighter and the toy comes through
as a tad darker than in the original images. That's just the way these things happen to combine.
Now, the next mode is called the divide mode and this takes the values from the bottom layer,
multiplies them by 256, then divides by the values from the bottom layer, but you add one.
And the reason you do that, what if you had a value from the bottom layer that was zero?
You know, pure lack of color in that channel. Well, you cannot divide by zero.
Right? That is, in mathematics, that's considered a forbidden operation, undefined.
So to avoid that, you add one to whatever the value is in the top layer.
So bottom layer times 256 and then divide that by top layer plus one and that gives you
your result. Now, we did it with the dog image on top and basically got something that was very,
very light. The dog image basically disappeared entirely from the combination.
The dark parts, particularly the parts that were similar in color to the dog,
have just changed color entirely. You can kind of make out a little bit of the dog in like the
body of the wizard and stuff like that. Well, this is exactly the case that we talked about when
we put boundary conditions on this. The toy image on the bottom is very light. The dog image on
the top is fairly dark. So the result of combining them with divide therefore produces a very
washed out image. Now, reversing the layers and putting the toy on top is very different.
Now, the places where the dog is up against the white background of the toy image,
well, the dog just comes through virtually unchanged. For the toy, the areas that were dark
when the dog was on top are now lighter. In fact, pretty much everything is a little bit lighter
except the base of the crystal ball kind of thing. Now, one interesting use of this mode is to
remove particular tones from an image. As an example, I took a color that is in the dog image.
It's kind of a mid-tone brown. Now, how do you take a color from an image? Well, there's something
called the color picker. It's the icon looks like an eyedropper or something like that.
You click on a pixel in the image and it will pick up whatever that color is. You take a look
in the color picker for the background and foreground and it will give you the numbers.
So this particular one was 857369. Okay, so a little more red than blue, but you know,
it's a mid-tone brown so you can plug that in and see what that looks like. I created a layer
with that color. I put that layer on top and the dog layer on the bottom and then use divide mode.
And what happens is there are pixels that have color in the original image
that have basically just gone white. So it's an interesting effect that removes that particular color
out of the image. So any pixel that had that color gone now.
So those are the inversion modes. We're getting near the end. There's only two more groups to go.
So we'll pick that up next time and for now this is a hook up for hacker public radio signing off
and as always encouraging you to support free software. Bye bye.
You've been listening to hacker public radio at hackerpublicradio.org.
We are a community podcast network that releases shows every weekday, Monday through Friday.
Today's show, like all our shows, was contributed by an HPR listener like yourself.
If you ever thought of recording a podcast and click on our contributing to find out how easy it
really is. Hacker public radio was founded by the digital dog pound and the infonomicon computer club
and is part of the binary revolution at binrev.com. If you have comments on today's show,
please email the host directly, leave a comment on the website or record a follow-up episode yourself.
Unless otherwise stated, today's show is released under Creative Commons
Attribution, share a light 3.0 license.