What type of pixel interpolation does Transform_World use? (SDK)

[Richard Rosenman]

Hi gang;

I am developing a plug-in that requires heavy downscaling without using Transform_World. My tests with bilinear are not good enough (leads to heavy aliasing) and bicubic is quite slow.

So I was wondering if anyone knew what type of pixel interpolation Transform_world uses? The results are quite good and I'd like to apply a similar approach.

Thanks in advance,

Richard

[shachar carmi]

from what i saw, it uses bilinear in HQ and nearest neighbor in LQ.

perhpas it only uses bilinear for upscaling, and uses a mipmap for downscaling.

[Richard Rosenman]

Hi Shachar;

Thanks for your reply. Yeah, that makes sense.

I'm pretty sure it doesn't use bilinear for downscaling because in my own test using standard bilinear scaling a 512x512 graphic down to 15x15, for example, results in brutal aliasing. Whereas the After Effects transform_world result is very good. So it must be using a different type of interpolation.

-Rich