On generating ground-truth time-lapse image sequences and flow fields