Masking questions

[Marpel]

Couple questions regarding the Masking Tool.

Before that, though, I should say the new masking tool/capabilities are great.

First is more a general question that relates to other operations as well. What impacts the Cancel Button on transforms? In the past, I have clicked that button to stop an operation and the operation continued on for some time without de-activating the operation. Today, I tried to cancel a zoom transform that was suggesting over 5 minutes to complete. When I hit the cancel button (in both places and a couple times), the operation continued until the clock ran down and then the operation stopped (with no completed operation of course). Is this just a computer capability issue or other?

Second, If I do a composite with a mask (rectangle tool), and complete that composite, then do another composite, intending to do a variation of that first mask with the upper left point at exactly the same point but the rectangle a bit smaller (by moving the bottom right bounding box towards the upper left point), I can't figure out (if even possible) how to get the bounding box to return. I have tried "Reload last settings" in both the main dialogue (which just gives me a complete black mask) and the Base Mask dialogue (which seems to be always greyed out regardless of the operation). Any suggestions?

Third, is there any way to generate a mask, using the spline tool, that has a much larger feather radius (64) than is already provided? I am wishing to generate a very large feather-off of a composite effect. The Blur tool will provide a large feathering but it, of course, feathers in both directions, which I don't want. I have tried the Oval Gradient with unsatisfactory results so far.

Marv

[jsachs]

On Cancel, when I try it here is seems to reliably cancel the second time I click the Cancel button. Not sure why it takes two times.

On Composite, there are a couple of ways I can think of to do what you want, but they both work with the same idea.

1) The most general way is to save the mask settings from the first mask in a settings file and load them into the second mask. Then the Undo button should let you edit the rectangle. To save the mask settings from the first Composite, bring up the Settings menu in its Mask dialog box (not the one in the transformation dialog box). Then create the second Composite transformation, make an empty mask with New Mask... and then load the saved mask settings file.

2) Since both transformations are Composites there is a simpler way -- you can just duplicate the first Composite (Select it, then Ctrl-Click on it, then click the Copy button in the image browser). Then double click on the copy, re-open the mask dialog box, and click the Mask Undo button to let you edit the rectangle size.

On Feather, one way is to feather the first time with Falloff set to Step which will enlarge the mask, then feather it a second time with normal Falloff to soften the edges. In the meantime, I increased the maximum feather radius to 256 pixels for the next release and made the slider nonlinear so it is still possible to set small values easily.

[Robert Schleif]

As Marpel mentions, at least some transformations seem to ignore the cancel button until they are finished. Precise Gaussian is one where stopping the transformation before it completes its task would be useful. Could it check for depression of the cancel button once every twenty seconds or so? Every once in a while I mistakenly start the transformation with too large a blur radius and then wait and wait and wait before I can proceed to do anything.

[jsachs]

Precise Gaussian on my computer, like Zoom Blur, cancels reliably on the second click of the Cancel button. The transformations in question already check for Cancel each time they advance the progress indicator to indicate that one more row of the output image has been processed. Strips of rows are processed in parallel using one processor thread per strip -- usually there are two threads per core. How Windows rotates execution between the threads running the transformation and the threads running Windows is an opaque mystery.