A couple of mask questions

[MarkT]

Firstly, let me congratulate Jon and Kiril for continuing to make PWP such a great tool for photographers. Version 7 adds some really cool functions and I look forward to getting familiar with it all.

Now a couple of questions about masks. In particular I was very interested in seeing the "Precise Gaussian" blur method in action. I've noticed that when I build a mask selecting the foreground or in-focus subject area, then apply the P-G blur, of course the selected area gets the blur effect, so I switch the Amount sliders in the transformation window. However this results in a weird overlapped fringe. If I return the sliders, then Invert the mask, the blur effect is then applied correctly to the background portion of the image. Can the difference between the Invert option (mask) and switching the Amount sliders (transformation) be explained?

And here is a suggested enhancement. The new Watercolour effect works beautifully. Is it possible to add the ability to use a mask in the transformation?

(oh, and a side note: The Conformal Mapping window now displays correctly. Thanks!)

[Marpel]

Mark7,

As it was explained to me, you must ensure the mask is the right way prior to selecting in this particular transform (i.e. the item you wish to remain clear must be covered by the black part of the mask and the intended blurred section is covered by the white part of the mask). The mask cannot be inverted by using the slider as it can in other transform operations and, as you have found, trying to invert the mask by reversing the slider introduces that funny halo. I found I had to generate a mask as usual, then invert it before bringing it into this operation.

Marv

[jsachs]

The next upcoming beta version will have a separate control for specifying the blurring mask as distinct from the Amount mask. The blurring mask is used solely to specify what parts of the image to blur. The Amount mask - as with the other blur options - is used to selectively blend the input image with the blurred image. Hopefully this will make the process of specifying the mask less confusing as several others have run into the problem you mention already.

[jsachs]

For various reasons, transformations for which the result image is a different size from the input image do not have an Amount control.

[MarkT]

Thank you Jonathan, the changes you have outlined sound perfect (and thanks Marv, I guess I just found the use of the sliders a little confusing at first...). This will become an often-used transformation, I'm sure.

I didn't think about the image size difference in the Watercolour function. Makes sense, and I can always use a Composite to blend in the effect later.

[Robert Schleif]

It sounds like Precise Gaussian can be adjusted so as not to sample from pixels under a mask. Do I understand correctly? This seems to be the behavior that several years ago was noted as something that would be very useful when blurring backgrounds (Unfortunately, the search function on the message board doesn't seem to find that post.). At the time Jonathan noted that a transformation with these properties would be very slow, and is apparently the reason the blur radius in Precise Gaussian is limited to 16 pixels. If there were a need for blurring with radii larger than 16, significant speed gains could be achieved by not using every (unmasked) neighboring pixel when computing a blurred pixel.

[jsachs]

Precise Gaussian already bypasses blurring for pixels where the mask is black.

I did increase the maximum blur to 32 pixels for the next release.