AntiAlias 0 not working properly

[Sug]

The AntiAlias = 0 has been set in a image meter and the scaling is still using antialiasing.
Original image:


Rainmeter skin:

Code: Select all

[Rainmeter]
Update=50

[MeterOneIcon]
Meter = Image
ImageName = #@#\ico.png
AntiAlias = 0
W = 256
H = 256

Result:


Is this a bug or is there a catch?

[jsmorley]

Rainmeter is just not going to successfully resize a 16x16 image to 256x256. Nothing will really.

[Sug]

Rainmeter is just not going to successfully resize a 16x16 image to 256x256. Nothing will really.

Not sure what software you use, but most stuff successfully resize images.
I'm not talking about the "quality" of the result, I'm talking about the antialiasing in the result.
A result without antialiasing should look like this:

Not like what was given.

[jsmorley]

Not sure what software you use, but most stuff successfully resize images.
I'm not talking about the "quality" of the result, I'm talking about the antialiasing in the result.
A result without antialiasing should look like this:

Not like what was given.

Right, I was talking about the "quality" when you resize such a small image. In any case, Rainmeter is just not good at that kind of resizing. I do think that antialiasing is unavoidable with images when resizing "up" with the method Rainmeter uses. I would really suggest resizing it in something else, and saving the larger file, then just use it as-is in Rainmeter.

[Sug]

Right, I was talking about the "quality" when you resize such a small image. In any case, Rainmeter is just not good at that kind of resizing. I do think that antialiasing is unavoidable with images when resizing "up" with the method Rainmeter uses. I would really suggest resizing it in something else, and saving the larger file, then just use it as-is in Rainmeter.

I see.
Appreciated.

[jsmorley]

I'd be interested in what software you used to scale up that 16x16 png image without getting "blurring".

If I download that hXB8u7B.png image from your first post, everything I use to zoom it to 256x256 pretty much looks like this:

1.jpg

I tried Photoshop and FastStone Image Viewer, both the same...

Edit: Ah, ok. I was able to get that by zooming with a "nearest neighbor" resize method. Yeah, not supported in Rainmeter.

2.jpg

[Yincognito]

Edit: Ah, ok. I was able to get that by zooming with a "nearest neighbor" resize method. Yeah, not supported in Rainmeter.

Yep, that was it. By the way, does Rainmeter have a method to do this or it's just some method baked in the functions you use in the code itself?

[Brian]

When we switched over to D2D rendering, we had to make a judgement call when rendering images and shapes.

In D2D, drawn images basically ignores anti-aliasing altogether and relies on a defined interpolation. Instead of using the nearest neighbor method, we chose the high quality cubic method which basically pre-downscales the image (if needed) then applies a cubic interpolation. This produces much better image quality. We planned on introducing an "interpolation" option for images at some point, but it has not come to fruition.

For the Shape meter, aliasing on vector graphics just looks terrible, especially when using transforms, so we chose to ignore the AntiAlias option altogether. I guess we could just default AntiAlias to 1 in this case, but I am not sure if that is very consistent with other AntiAlias options.

-Brian

[Yincognito]

When we switched over to D2D rendering, we had to make a judgement call when rendering images and shapes.

In D2D, drawn images basically ignores anti-aliasing altogether and relies on a defined interpolation. Instead of using the nearest neighbor method, we chose the high quality cubic method which basically pre-downscales the image (if needed) then applies a cubic interpolation. This produces much better image quality. We planned on introducing an "interpolation" option for images at some point, but it has not come to fruition.

For the Shape meter, aliasing on vector graphics just looks terrible, especially when using transforms, so we chose to ignore the AntiAlias option altogether. I guess we could just default AntiAlias to 1 in this case, but I am not sure if that is very consistent with other AntiAlias options.

-Brian

Thanks for the info, Brian.