Handle content with > 8 bit

[slhck]

Suggestions on how to deal with content that is > 8 bit:

Option 1:

• Truncate to 8-bit

Option 2:

• Convert input to 0–255 range, keep it as float (just like for full range content)
• Convert using regular functions

Option 3:

• Keep the input as it is, use the functions as-is
• Convert back into the regular SI scale for 8-bit by
  • dividing by difference between max and min value (max - min + 1)
  • based on the theorettical range of the input (e.g. 235 - 16 + 1)

Group decision seems to go towards option 3.

[slhck]

Not quite sure what to do with SDR, 10-bit. Need to adapt the CLI interface a little bit, or infer the bit depth from FFmpeg.

[slhck]

Now we're actually handling all content between 0 and 1 and apply the HDR conversions functions there. Then we still have output in the range of 0-1.

This means SI/TI are calculated on that range, and not the original range.

[slhck]

Solution is now implemented to convert back to original range 8-bit range.

See: https://docs.google.com/presentation/d/1zvdbEOmnBWbSE3YrI9VWzpZO35VAAmO8CQ33URRkq9k/edit#slide=id.gd0fe7ad648_0_5

[slhck]

This should be fine now as-is with the scaling to 0-1 and later upscaling to the 0-255 range.