diff options
author | Søren Sandmann Pedersen <soren.sandmann@gmail.com> | 2020-01-29 23:25:11 -0500 |
---|---|---|
committer | Søren Sandmann Pedersen <soren.sandmann@gmail.com> | 2020-01-29 23:25:11 -0500 |
commit | 1e35f9892f4b46163637ca30e2275ec5490ce692 (patch) | |
tree | b04c93c023c8b104750ea1fe17792e22bf4c799a | |
parent | e5c864b7a85d67eae3dd3956df9149970c60b0b3 (diff) |
asdfdocs
-rw-r--r-- | docs/deviation.txt | 14 |
1 files changed, 14 insertions, 0 deletions
diff --git a/docs/deviation.txt b/docs/deviation.txt index 42f38693..19046f08 100644 --- a/docs/deviation.txt +++ b/docs/deviation.txt @@ -43,3 +43,17 @@ we get A reasonable value for acceptable_error might be 2.5% or 3%. For a 1 bit format 2.5% is a little less generous than the current 0.0064 (it is 0.0062), whereas 3% is a litt more generous (it is 0.0074) + +Another, perhaps better, possibility is to just reason as follows: + +By choosing a particular format, the user has indicated that they +find an error of half a step in that format to be acceptable. We could +simply say that pixman guarantees that the error is always less than +say 52.5% of a pixel step. + +So a true value of 0.525 would be allowed to result in both 0 and +1. But anything bigger than that would have to be 1. + +The more complicated scheme has the advantage that it doesn't refer to +pixel steps and so the errors could be computed in some perceptually +uniform color space such as CIELAB. |