Sorry, Just realized you responded -

I’m not sure if your other thread takes it into account or not. I was meaning a video scanning technique - here’s a simple example to try and explain:

Frame 1: 7 shades of blue, 6 shades of green, 3 shades of red
Frame 2: 3 shades of blue, 3 shades of green, 1 shade of red
Frame 3: 4 shades of blue, 5 shades of green, 4 shades of red
Frame 4: 4 shades of blue, 4 shades of green, 4 shades of red
Frame 5: 5 shades of blue, 3 shades of green, 5 shades of red

My idea is to scan the file for the frame with the most colors (with a small bias of evenly dispersed colors).
In this case the scan would get a result of “Frame 3”, since it has the highest minimum color, as well as an extra shade of green. Even though “Frame 1” technically has the highest color variance, it is not chosen because of the lower minimum for shades of red - this avoids radically over-defining some colors at the expense of others. However, “Frame 4” is also passed by because after the minimum has been established, the additional shade of green in “Frame 3” is still helpful in defining the color shift.

Since the Color Correction model is based on a single frame matched between two sources, wouldn’t this help identify the frame most useful for the process?