Q&A about *NON* Math components

I believe the XY anchor is a sort of "pivot point". If you wanted to rotate your image not around the exact center, but based on the upper left corner you might put X=0 Y=0

Coordinate system is like this

(X=0,Y=0)
--------------------> + X
|
|
|
|
| `````````` (X=.5,Y=.5) <-- Default Center of rotation
|
|
|
\/
+Y

Try modifying the origin x, y on the rotate component, then fiddle with the "Rotation" wheel. It should be fairly intuitive after that. I'd recommend keeping the origin values fairly low, else the rotate will start to act all crazy.

The XY Anchor input is present in components that allow to specify point coordinates. It determines how the coordinate values are interpreted. It affects coordinates of all points in a component.

Default (Size)

This option makes all points in the component change their positions together with the changes of the Size slider. Say, if you blend a Free Ellipse with XY Anchor set to Default(Size) with a Perlin Noise and then change the Size slider, the ellipse will 'stick' to the noise -- they will scale together, there will be no discrepancy between them as the Size value changes.

In this mode, a point with coordinates of 0, 0 refers to the upper-left corner of the image, and a point with coordinates of 1, 1 refers to the lower-right corner of the rectangle Size x Size pixels wide, where Size is the current value of the Size slider.


Max Size

The idea behind this mode is that it makes point coordinates independent of the Size slider -- just like the Image Bounds mode described below -- while avoiding its inherent problem with non-equal distances along coordinates that may arise due to non-square images.

In this mode, a point with coordinates of 0, 0 refers to the upper-left corner of the image, and a point with coordinates of 1, 1 refers to the lower-right corner of the rectangle MaxSize x MaxSize pixels wide, where MaxSize is the maximum value of the Size slider (you can see the maximum Size value by setting the Size slider to the rightmost position.)


Image Bounds

This option makes all points in the component independent of the Size slider. Point coordinates are interpreted as percentages of width and height of the source image loaded into Filter Forge.

In this mode, a point with coordinates of 0, 0 refers to the upper-left corner of the image, and a point with coordinates of 1, 1 refers to the lower-right corner of the image, no matter if the image is square or rectangular.

Here's how the filter looks from the inside. Note that previews on all gradient component, and the final RGB assembly look identical -- this is because the previews are square and Size is set to its maximum value:

And since I was only working on square Images, I saw no result. OK, that sounds logical Thanks Vlad

Nahhh, I'm just gonna stay square !

Quote
Sphinx. wrote: It seems to ignore many intermediate values and suddenly "jump" to a higher radius..

Haven't checked the code, but the jump you're seeing might be caused by the algrothm itself -- these components look at all pixels within the radius and choose one pixel that has the "most" minimum/maximim/median value. Even if you have a floating-point radius, just slightly overlapping it with a black pixel is enough for black to "take over" the output pixel.

(The anti-aliasing option softens that effect slightly -- it calculates four median/minimum/maximum pixels and averages them to produce the output pixel. Technically, this is not an entirely correct approach, but it looks better than the jaggy correct version.)