Selection Width, Height and Center Coordinates

The selection must be same size as the image loaded. If you use Transform > Scale component, just set the XY Anchor to "Image Bounds", then 0.5,0.5 will be the center, and you can set any scale you want..

Oh! That's a whole different story!

There is no simple way to do that, because you only have information about the whole selection map as such available. Selection data can also be gradual (i.e. graytones).

What you need here is some sort of autotrim method that gives you the bounding rect of non zero selection samples. This is somewhat complicated to figure out when you work in a procedural rendering context..

Maybe it would be better to go at the problem from another perspective: what is your overall idea / goal here?

Yes, a script is currently the only way to search down the bounding rect, and it can be done in a fairly optimal way IF we just didn't have that darn initialization overhead issue...

But even with that problem, you could optimize the search using this method.

Basically you perform the search inside the initialization section. The search must be performed in two steps:

1. search from top,left and forth to first encounter of a non zero sample (or the sample value of the top, left coordinate). At first encounter break the loop.

2. like above, but search from bottom, right and back instead.

you can output the values through r,g,b,a (top, left, bottom, right).

Btw..alternatively to your blur attempt, try using maximum or median instead of a blur...

Sphinx, thanks anyway.
I will try these approachs again until another approach will be found.

ThreeDee, thank you for your reply.

Quote

ThreeDee wrote: A thought on this matter: Not the prettiest of solutions, but you could blur the selection image with directional blur horizontally for the vertical values and vertically for horizontal values (with motion blur component set to radius 100, directional 100, angle 0 or 90, not gaussian). Then perform the seach for non-zero values along x and y-axis from the edges on in. This way you'd only need to seach one row of pixels in each direction. I assume this would be faster than a purely script-based solution.

You're amazing !
Now I can get the value of selection area width, height and center coordinates.

I used "Motion" blur instead of "Directional" one.
Because there is no "Directional Blur" Component.
Except it, I used your method.

If selection area has very long edge, it may need more steps below.
"Directional Blur" >>> "Threshold" >>> "Directional Blur"...

I'm going to continue to improve. Thanks.