Improvements to grain synth gamma handling #622
Conversation
av1an-core/src/grain.rs
Outdated
| TransferFunction::SMPTE2084 => { | ||
| if x == 0.0 { |
There was a problem hiding this comment.
Floating point comparisons to 0 should probably be done against an epsilon, like x.abs() < f32::EPSILON
| TransferFunction::BT1886 => 0.18, | ||
| TransferFunction::SMPTE2084 => 26. / 10000., | ||
| } | ||
| self.to_linear(0.5) |
There was a problem hiding this comment.
This seems like it should call from_linear (since from_linear takes a value between 0 and 1), but I'm not sure. Could you explain why to_linear is called here?
There was a problem hiding this comment.
The midtone we're seeking here is the perceptual gray midtone according to the transfer function, i.e. based on this transfer function, what value with the transfer function applied appears to be "half brightness". This is calculated by doing the transfer calculation on a mathematically half luma value, i.e. to_linear(0.5), and for e.g. a flat gamma curve of 2.4, results in a midtone of about 0.189 (approximately the previous constant value).
- Fixes BT.1886 to use the more correct formula and gamma - Simplify the `mid_tone` function - Add quickcheck tests around `to_linear`/`from_linear`
shssoichiro
force-pushed
the
bt1886-grain-updates
branch
from
2edb257
to
2acad1f
Compare
mid_tonefunctionto_linear/from_linear