Hi. I'm running the Compute Difference method on an island scene. All I really want to do is derive two change classes from one scene in 1988 to another in 2007: vegetation to vegetation (little or no change beyond seasonal, which are minimal as the location is tropical) and vegetation to developed. I've already georeferenced them and performed a dark object subtraction to minimize atmospheric effects. I would like to be able to eliminate the surrounding ocean water from the analysis but do not see anywhere I can do so -- the same goes for a cloud mask I created when running a separate change detection analysis based on maximum likelihood. I think the presence of water, even after spatially subsetting as much as possible without excluding terrestrial areas, is serving to skew the Compute Difference results. Also, it makes for extremely difficult visual interpretation. I already have the water and cloud areas as ROIs. Is there a way to eliminate them from the process? All I'm interested in is the two types of land cover change (zero to minimal and veg to developed), but between trying to arrive at the proper number of thresholds, threshold ranges and normalization vs. unit variances, I'm being thwarted by the presence of all the water. Also, is there a way to examine the normalized (or unit variance-adjusted) images before the difference map is produced? Thanks for your assistance! Mike