The resample kernel is a decimation resampling filter used to resample
dense data. It works on the keyframes and may not produce the desired results
when used with sparse data.
The resample filter converts from either uniform or non-uniform
timestep input data samples to the specified uniform timeStep. Various
time domain filters are available and are specified with the kernel
flag which selects the resampling kernel applied to the keyframes
on the animation curves.
Kernel Values
closest |
Closest sample to output timestamp |
lirp |
Linear interpolation between closest samples |
box |
Box filter: moving average |
triangle |
Triangle filter: (1 - |x|) weighted moving average |
gaussian2 |
Gaussian2 Filter: (2^(-2x*x)) weighted moving average |
gaussian4 |
Gaussian4 Filter: (2^(-4x*x)) weighted moving average |
This filter is only targeted at decimation resampling
-- interpolation resampling is basically unsupported. If your
output framerate is much higher than your input frame rate
(approximate, as the input timestep is not assumed to be regular)
the lirp and triangle will interpolate (usually) and the rest will
either average, or use the closest sample (depending on the
phase and frequency of the input). However this mode of operation
may not give the expected result.
|