Q:Why inverting a feathered mask isnt pixel perfect?

[Miguel_Curto]

Ok, this one had me puzzled for a long long time, and I've accepted and found work-arounds to cope with it but its something I quite can't still understand why.

When one inverts a feathered mask there's always some pixel gap between that one and previous and never understood why.

Like the example If I put a black layer with a feathered (say 1px) mask and duplicate that layer and invert mask, I think its safe to assume one would expect a full black rendering, but it isn't so.

Any ideas why?

Ty

[Barbara Ash]

The mask is feathered so the edge pixels are not fully masked in the feathered area. That's the nature of feathered masks. It's working as expected. If you don't want a gap, avoid feathering and use a hard-edge mask.

[Miguel_Curto]

I'm sorry but I need a little more elaborate than "nature of it", from my view any thing inverted should be "Absolute" inverted.

Any feathered mask is a ramp from black to white in mask, from I gather the two grey areas (50% ramp) should coincide, but they dont.Could be to the number of steps being a odd number (as opposed to even), I really don't know so was hoping for a bit more technical reason why.

I do need  feathered masks frequently and need to work with the inverted ones, my work around usually is to apply some more feather than needed and use Levels on both to displace them a bit.There can be other work-arounds thought, that's why I'm asking.

[Barbara Ash]

Feathering is a radius. With a 1 pixel feather, you actually have 2 pixels, one of each side of the edge. Hope that helps.

[D Fosse]

I have a hunch (not tested) that it's because of grayscale gamma encoding. The mask is a grayscale image, as per your working gray.

[Miguel_Curto]

I managed to spend some time testing and I'm fairly certain its not Gray Gamma settings, still don't know but the closest I could get is if you turn off "Blend RGB Colors Using Gamma X (where X is 1,00 as default) and use Color Space Gamma instead (mine being Adobe RGB that uses 2,2) the gap shrinks, doesn't'completely disappears but its noticeable smaller.

The "remaining" could boil down to anti-aliasing, don't know...