Image Processing & Correction
Scaling factor
A meta-parameter — the same dial in many engines, with the same purpose: keep results in the bit-depth range you have
Definition
A scaling factor is a numeric multiplier applied to an engine's computed output before it's written to the result image. If an engine computes some intermediate value v per pixel and the scaling factor is s, the actual output is s · v. The role: bring the computed intermediate into the bit-depth range of the output. Two images added together can produce values above the 8-bit ceiling; a scaling factor of 0.5 brings the sum back into range. A polynomial of an image can produce tiny fractional results; a scaling factor of 100 brings them into a useful range. The engines referenceLoading... names which engines accept the parameter; the rule of thumb across all of them is the same.
It's a gain on the output
Multiply the result before writing it
The default may not be the right value
Whether to scale depends on the math the engine just did
Rule of thumb: aim for the histogram to fill the range
Scale until the output uses the full bit depth
32-bit intermediates avoid the question
If you don't downcast, you don't need to scale