[Bug] Antialiasing difference when applying an image mask

[Cariboudjan]

Untitled.png

When an image receives an image mask it decreases in size by 2 pixels. Appears to be antialiasing the image resulting in the appearance of a smaller meter, rather than actually changing the image's dimensions.

[jsmorley]

Untitled.png

When an image receives an image mask it decreases in size by 2 pixels. Appears to be antialiasing the image resulting in the appearance of a smaller meter, rather than actually changing the image's dimensions.

I would be tempted to use Container rather than MaskImage. I really think Container will do all you want, and it shouldn't have any weirdness with how it renders images. MaskImage is in effect deprecated, and I don't recommend it.

[Cariboudjan]

Would love to but containers actually do not work in this specific application, since the meter itself is also inside its own container, and containers cant exist inside another container.

[jsmorley]

Well, the bad news is that we are just done with MaskImage, and I don't foresee any changes being made to it. You might want to re-think the approach are using if MaskImage isn't working well for you.

[Brian]

I wish we could fix this, but this is a limitation of D2D. The masking requires "aliasing".

https://docs.microsoft.com/en-us/windows/win32/api/d2d1/nf-d2d1-id2d1rendertarget-fillopacitymask(id2d1bitmap_id2d1brush_d2d1_opacity_mask_content_constd2d1_rect_f__constd2d1_rect_f_)#remarks

And here is where we deal with it:
https://github.com/rainmeter/rainmeter/blob/master/Common/Gfx/Canvas.cpp#L624-L647

-Brian

[Yincognito]

Just speculating here, but maybe you could apply a TransformationMatrix to the image to make sure it stays the same size visually, then use a partially opaque padding approach of some kind (if possible) for the underlying meter to make sure the clickable area stays the same too...

[Cariboudjan]

It's not a big deal. I just thought I'd share. It's not breaking anything.