Ronounours wrote:
@PC : Basically, this is a similar implementation to the plug-in 'Colorization', already discussed here :
viewtopic.php?f=9&t=839&start=0.
But as the Colorization plug-in is not maintained anymore, I think this is cool to have such a feature in G'MIC (especially as it will be easy to maintain, it took only about 10 lines of G'MIC code :p).
yes a very nice idea indeed
Quote:
The color extrapolation works by extrapolating the chromaticity values of the colors in Lab. I've some ideas for improvement :
- Ability to select the output mode, either as a single RGB image (as it is done now), either as two layers that should be blended in 'Color' or 'Value' mode, so that we can correct the color extrapolation on some regions where artefacts remain.
- Ability to select amongst several color spaces for the extrapolation : either Lab (as it is done now), but also YCbCr and RGB, with or without keeping a strict lightness constraint.
Maybe I'll have time to do this tomorrow.
I really like the results ( i make a quick confront with colorize..nothing worth to be posted but result were good
Quote:
- Ability to select the output mode, either as a single RGB image (as it is done now), either as two layers that should be blended in 'Color' or 'Value' mode, so that we can correct the color extrapolation on some regions where artefacts remain.
That would be perfect for what is spinning in my mind
I never found written references for this but i get often result i like more by
1 )Slightly blurring the "color" layer (except with unreasonable dramatically high values this do not make the resulting image less sharp, but somehow make more clean )
2 )Slightly sharpening and darkening the grayscale layer
(this instead impact a lot on result sharpness )
i do this almost always when using value or color modes,
not always but often i also create additional dup of the layer used , and i add to the visible once merged with low opacity in soft light or overlay mode (to make colors more brilliant and/or increase shadows/highlights contrast )
Again i have no references for this... just my own experiments