Have you tried both methods to compare, Ken?
My preference would be to process at the native resolution and resample as a final step. The logic is that any "extra" noise precipitated from potentially finer enhancement is somewhat tempered by the subsequent sampling (through loss of resolution!)
|