Hi Makr, I would use the 'Band Threshold to ROI' in the ROI tool to do this. For each unmixing result, find a threshold for the dominiant endmember (for example, any pixel with a value greater than 0.6) then use the Band Threshold to create an ROI of these pixels for each endmember from the LSU. Then, from the ROIs, you can create a classification image. Alternately, you could create a Band Math function that would do something similar. -mm