Okay more detail on the problem at hand. I've used some filters to isolate clouds and have replaced the cloud pixels with pixels from an overlaping ASTER granule. Some of the cloud pixels were replaced by other cloud pixels, but much of the cloud cover was succesfully replaced by non-cloud pixels. The bands correspond between images, of course, and they have all been processed in exactly the same way. Moreover, the granules I'm using were captured by ASTER within the same season at nearly the same time of day (though in different years). Atmospheric correction and cross-talk have been corrected for (by people at NASA or the Japanese lab since I simply took the pre-processed data). So, what's the problem then? After patching the images together, I noticed that there was an obvious difference between the patched areas and the surrounding areas. When I use ENVIs handy profile tool, the break in the profile is also obvious were the patches are. When I go to classify this pathced image, those areas are going to come out as distinct from the surrounding areas even though the actually represent the same biogeographical features. I can think of a few time consuming ways to adjust the patched areas so that they correspond more closely with the surrounding terrain -- ways that I'm certain I can justify -- but I'm hoping that there is a faster solution in ENVIs tool kit for dealing with the problem. Has anyone else encountered this difference between images of the same band in ASTER data and (given that nearly all the convievable variables related to the data gathering - time of day, season, etc) what could be causing the discrepency? Thanks, Chris