ENVIers: When resizing or reprojecting classification rasters, many times I would prefer to take the most common class falling within the output pixel, not the nearest neighbor. Is there any trick to doing this with ENVI? Bilinear and cubic convolution are not appropriate with categorical data, which is why I ask. The only way I could think of would involve making a single image for each class filled with 0s (where the class is not) and 1s (where the class is), do the resample where the output is the % cover by that class, and then determine the max cover across all classes. This seems exceedingly complicated to do without batch-scripting it. Is there any other way? -j