FerretBomb
Active Member
Not sure if this would be entirely feasible, but would it be possible to add a chromakey option to make getting the correct 'target' value easier, with shadowing variances?
Thought was to take 3-4 eyedropper samples (bright spot, two mediums and a shadowed spot), and use them to plot a chroma 'vector' where all of the samples are present (if possible) on a single luminance/brightness curve, then allow two settings (rather than just 'similarity'); one for 'hue similarity' and one for 'luminance similarity'. Essentially making an 'oval(/ellipsoid)' target for the key, rather than flat circular (as it appears to be now).
Would allow shadowed and bright spots to be keyed still, while keeping the hue range tight to prevent keying out other colors as well (bleed). Would also help those using posterboard and green duct tape as a makeshift screen, to really dial it in tight.
I understand that lighting the chromakey screen/backdrop properly is greatly preferred/the right way to handle this issue. But this would add a lot of flexibility and a significant feature-add for OBS, if it would be possible.
If the chroma-vector idea wouldn't work, averaging multiple samples would still help grab a mean value and make the best basic(/circular) 'similarity' value easier to hit.
Thought was to take 3-4 eyedropper samples (bright spot, two mediums and a shadowed spot), and use them to plot a chroma 'vector' where all of the samples are present (if possible) on a single luminance/brightness curve, then allow two settings (rather than just 'similarity'); one for 'hue similarity' and one for 'luminance similarity'. Essentially making an 'oval(/ellipsoid)' target for the key, rather than flat circular (as it appears to be now).
Would allow shadowed and bright spots to be keyed still, while keeping the hue range tight to prevent keying out other colors as well (bleed). Would also help those using posterboard and green duct tape as a makeshift screen, to really dial it in tight.
I understand that lighting the chromakey screen/backdrop properly is greatly preferred/the right way to handle this issue. But this would add a lot of flexibility and a significant feature-add for OBS, if it would be possible.
If the chroma-vector idea wouldn't work, averaging multiple samples would still help grab a mean value and make the best basic(/circular) 'similarity' value easier to hit.