Closed johnnychen94 closed 2 years ago
Merging #72 (fb22593) into master (7969627) will decrease coverage by
0.08%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## master #72 +/- ##
==========================================
- Coverage 98.44% 98.36% -0.09%
==========================================
Files 14 14
Lines 257 244 -13
==========================================
- Hits 253 240 -13
Misses 4 4
Impacted Files | Coverage Δ | |
---|---|---|
src/DitherPunk.jl | 100.00% <ø> (ø) |
|
src/error_diffusion.jl | 100.00% <100.00%> (ø) |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 7969627...fb22593. Read the comment docs.
I'm surprised that this actually makes things slower...
Edit:
This might be implementation-specific; if the filter weights are StaticMatrix
then the performance regression here can be eliminated. I'll try to rework the error diffusion dither implementation on top of this PR and see if we get better results.
This might be implementation-specific; if the filter weights are
StaticMatrix
then the performance regression here can be eliminated. I'll try to rework the error diffusion dither implementation on top of this PR and see if we get better results.
I've had some shower-thoughts about this. I think we should fully drop the dependency on OffsetArrays
and use CartesianIndices to iterate over the neighborhoods defined in the filters. This might also simplify a future implementation of #20.
Closing this as it is outdated by #73 , which contains parts of this PR.
This is not type inferrable and would thus hurt the performance.
Changing the order of parameters is technically breaking, but I doubt people outside are depending on it. Thus an appropriate deprecation of
color_space
keyword should make this a transparent improvement.