VEX's finput() function's document says:

https://www.sidefx.com/docs/houdini/vex/functions/finput.html [www.sidefx.com]

> Returns fully filtered pixel input.

What does this mean exactly? I understand how cinput (nearest point sampling) and binput (bilinear interpolation) work. But what's "fully filtered pixel input"? Which filter?

It determines the position of the four corners of the transformed output pixel in the input's space and averages all the input pixels that fall within that quadrilateral. It's good for cases where pixels are squished (deformation, scaling down a lot). It's also the slowest of the methods.

malexander
It determines the position of the four corners of the transformed output pixel in the input's space and averages all the input pixels that fall within that quadrilateral. It's good for cases where pixels are squished (deformation, scaling down a lot). It's also the slowest of the methods.

It still sounds a lot like bilinear interpolation to me... or it's actually just doing bilinear interpolation 4 times (for each corner) then average the results?

raincole

malexander
It determines the position of the four corners of the transformed output pixel in the input's space and averages all the input pixels that fall within that quadrilateral. It's good for cases where pixels are squished (deformation, scaling down a lot). It's also the slowest of the methods.

It still sounds a lot like bilinear interpolation to me... or it's actually just doing bilinear interpolation 4 times (for each corner) then average the results?

bilinear blends between four corner values, finput() blends between all of the values inside the box.

Ah I see. That makes sense. Thanks you two.