pub fn interpret(
image1: &Image,
image2: &Image,
source: ChangeSource<'_>,
) -> Result<InterpretResult, InterpretError>Expand description
Describe what changed between two images.
The single entry point: a pixel diff, a similarity metric and a caller passing its own boxes all come through here and get identical treatment, because the statistics are computed from the source pixels either way.
diff_count therefore means the same thing on every path — actually-changed
pixels, never map windows.