If I understand this correctly its like a mask between filters and the differential makes that differential much more noticeable? Wouldn't one of the challenge

If I understand this correctly its like a mask between filters and the differential makes that differential much more noticeable? Wouldn't one of the challenges be that the pictures have to be almost the exact same time with those filters so that they line up perfectly to provide the differential given the high resolution? Also thanks for the detailed response - their approach sounds like a smart solution minimizing unnecessary compute cost / algorithm scanning.