The idea is good, but I think the performance is not wink The way you propose needs memory for 4 more images instead of a 2 line vector calculation as proposed by bart. But thank you for thinking laugh