Parameterize the cubic kernel in 'bicubic_scale.effect' #7811
Unanswered
jukofyork
asked this question in
Development Questions
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
The current implementationin of 'bicubic_scale.effect' just uses a fixed parameter set of {B=0, C=0.75}.
The original paper suggests that "satisfactory" parameter sets all lie on the line B + 2C = 1, eg:
So if limited to a hard-coded (cardinal) parameter set, the Catmull–Rom filter: {B=0, C=0.5} would probably make more sense and result in less unnecessary ringing.
But it would also be nice to be able to use other parameter sets, such as the Mitchell–Netravali filter: {B=1/3, C=1/3), etc.
Looking at the code, the change looks fairly easy to make:
But this might need to be optimized by somebody who understands how the ".effect" code gets converted to the shader language(s).
Beta Was this translation helpful? Give feedback.
All reactions