YOUR ACCOUNT

Login or Register to post new topics or replies
byRo
an Englishman in Brazil

Posts: 138
Filters: 8
...not a curve?

I was pretty comfortable with the idea that grey things are numbers, green things are images and blue things are curves.
or, to Geekify, none, two and one dimensions

But then I try to figure out why curve op's have a green input.
Have a look at the attachment. The only information going to the gradient is a curve, but this little blue (curve) connection is carrying a whole greyscale image.

Vladimir? onyXMaster? How does this stuff work??

Rô

_________________________________
My favourite question is "Why?".
My second favourite is "Why not?"
  Details E-Mail
onyXMaster
Filter Forge, Inc.
Posts: 350
Actually, while map components (green) are functions of two parameters (x and y), curves may appear to be functions of one parameter (called internally t). But practically, curves are functions of three parameters (t, x, and y). Curves which have map inputs ant therefore produce different values depending not only on t but also on x and/or y are called internally "mapped curves".

So the curve value depends not only on the value it remaps, but also on the x and y coordinates. In your case, the resulting curve value will different for every sample, since the "percentage" value is taken for each sample.

Let's talk pixels instead of samples for simplicity: imagine that FF calculates the pixel X;Y of the "Profile Gradient" component.
The profile gradient itself is a basic linear gradient, so the base t value for the curve is calculated using basic linear formula and yields the value T. After that, the value is remapped (since there's a Profile curve different from Linear pseudo-curve). So the Blend CurveOp is sampled with t=T1, x=X, y=Y. The Blend itself, samples input curves using the same t, x and y and gets the T'(sourceA) and T'(sourceB) values. Also, it samples the percentage map with x=X and y=Y, uses it's grayscale output V to control percentage and performs a blend between T'(sourceA) and T'(sourceB) using V. As you see, in this example, the value of V varies with change of X and Y, so you'll get different blend percentage for different input image.

Hope that wasn't too technical smile:)
  Details E-Mail
onyXMaster
Filter Forge, Inc.
Posts: 350
P.S. The preview of the mapped curve shows a finite number of curve shape approximations, with varied mapped coordinates -- it cannot and should not be used as a "median/average/statistical curve" for the image, just serve as a general preview.
  Details E-Mail
byRo
an Englishman in Brazil

Posts: 138
Filters: 8
Quote
onyXMaster wrote:
Hope that wasn't too technical
No, I think that was exactly the answer I was looking for - just give me a few hours to digest, OK?

Rô
_________________________________
My favourite question is "Why?".
My second favourite is "Why not?"
  Details E-Mail
onyXMaster
Filter Forge, Inc.
Posts: 350
Sure, feel free to ask more if you need -- after all since we don't have detailed documentation, it's the only reliable way to get info (apart from experimenting of course smile;) ).
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3446
Filters: 55
Quote
byRo wrote:
But then I try to figure out why curve op's have a green input.


Ahhh, you've discovered mapped curves smile:D I'll try to "de-geekify" onyXMaster's explanation:

When none of the green parameters of a curve component are mapped, the curve is same for every area of the filter.

But when you connect a map component to one of the map inputs of a curve component, the curve can have different shape for every area of the filter.

I admit, the idea of mapped curves is quite hard to digest, but it can lead to quite interesting results. I'm glad you've found this smile;)
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3446
Filters: 55
Here's a good example:

  Details E-Mail
Kraellin
Kraellin

Posts: 12749
Filters: 99
i'd already discovered this from playing, but what limits it is that the blue curve outputs can only go into the blue inputs and if you use a tone curve then the curve only affects the tone and never the shape or anything else.

craig

If wishes were horses... there'd be a whole lot of horse crap to clean up!

Craig
  Details E-Mail
xirja
Idididoll Forcabbage

Posts: 1698
Filters: 8
Quote
But when you connect a map component to one of the map inputs of a curve component, the curve can have different shape for every area of the filter.

I admit, the idea of mapped curves is quite hard to digest, but it can lead to quite interesting results. I'm glad you've found this


So from the recent Mountains at Sunset by Kochubey comes this attached image.

Please explain and help me avoid the $300/hr. psychiatrist sir.

_____________________________________________________

http://web.archive.org/web/2021062908...rjadesign/
_____________________________________________________
  Details E-Mail
Ramlyn
Ramlyn

Posts: 2930
Filters: 691
Good question. I'm curious too.
  Details E-Mail
Betis
The Blacksmith

Posts: 1207
Filters: 76
The black and white values of the map turn into a different "slider" value at each sampled pixel.

In this example: Where the input is white, the bias curve has a high bias value, and the dark parts of the image change the curve (for those pixels only) to have a low bias.

The curve preview you see has several different values that it displays to give you an idea of how the curve looks a different parts of the image (they're just stacked layers to give you an idea, they don't correspond to actual pixels)

The reason the orange turns bright and dark is similar to how plugging in the orange to a Gamma node with the same black/white map would give you a similar effect (in fact I think the Gamma node is secretly a tone curve node with the gamma curve plugged in under the hood and only the Gamma parameter exposed). Kochubey decided to use a custom curve instead of a gamma curve, for whatever artistic reason. It might be fun to play around with this and plug in different curves like Noise, etc.
Roses are #FF0000
Violets are #0000FF
All my base are belong to you.
  Details E-Mail
xirja
Idididoll Forcabbage

Posts: 1698
Filters: 8
Quote
gamma curve plugged in under the hood


smile8) , ahhh that helps. I was thinking, a constant can't be a gradient! smile:?:
_____________________________________________________

http://web.archive.org/web/2021062908...rjadesign/
_____________________________________________________
  Details E-Mail
Ramlyn
Ramlyn

Posts: 2930
Filters: 691
Hummmm.... Interesting. Thanks Betis.

Changing Bias with other curves gives a wide variety of different results. They also don't look easy to control( at least for me ).
  Details E-Mail
xirja
Idididoll Forcabbage

Posts: 1698
Filters: 8
Truly, I thought I understood, but alas, a wrench in the works...

The results of the connected components should be the same AND what's with the B values in the first place! Argh! smile:cry:

MaptoCurvePuzzle.ffxml
_____________________________________________________

http://web.archive.org/web/2021062908...rjadesign/
_____________________________________________________
  Details E-Mail
Betis
The Blacksmith

Posts: 1207
Filters: 76
That is very strange indeed. If you check "Preserve Colors" it does stay the same though.
Roses are #FF0000
Violets are #0000FF
All my base are belong to you.
  Details E-Mail
Betis
The Blacksmith

Posts: 1207
Filters: 76
Not sure if you were just trying other things, but this is what I meant in my earlier comment:

Roses are #FF0000
Violets are #0000FF
All my base are belong to you.
  Details E-Mail
xirja
Idididoll Forcabbage

Posts: 1698
Filters: 8
Oh thank goodness. D'oh! Yeah I was toying about, and I roughly understood. That good old difference test has been handy many a time. smile:)

Maybe a tone curve without preserve color is the artistic reason Kochubey was hinting at?
_____________________________________________________

http://web.archive.org/web/2021062908...rjadesign/
_____________________________________________________
  Details E-Mail

Join Our Community!

Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!

33,719 Registered Users
+8 new in 7 days!

153,544 Posts
+13 new in 7 days!

15,348 Topics
+71 new in year!

Create an Account

Online Users Last minute:

21 unregistered users.