Can you calculate the statistics for the input file and post them? There was an issue I have seen before when the data has a very large range. It may help to rescale the data using something like Stretch Data.

Hi Mari, I think you're right that there's something weird with the input file I'm using. I calculated basic stats, and the results are:

Dims: Full Scene (5,171,936 points)

Basic Stats Min Max Mean Stdev
Band 1 ********************** 2.843009 ********************** **********************

Do you know why the only value to show up in the report is the max? When I mouse over the image, the pixed values range from -1 to 3 ish, with a few small "no data" areas.

I know there is an issue if you have some very tiny values and very large values. They could even be assigned NaN or Inf or something like that. Some programs can assign 'nodata' values on export to a very tiny, unlikely value so it is not confused as a real pixel value. Do you know what the range of your data should be? For example, if you know it should range from -1 to 1, then you could make a mask based on thresholds below -1 and above 1, then apply the mask where you can specify a value more within the natural range. That way as you try to correct for the striping, those bad values are not propagated through the data.

I'm not sure what's happening here. When I view the image in another program (ArcMap) the stats all look good (values range from approx. -3.5 to 3), then in Envi I get the stats I posted earlier, where the only value that shows up is the max.

Just to check, I created a mask in Envi that ranges from -3.5 to 3. The result was an image with every single pixel masked. So, even though the stats aren't showing up in Envi, all the values in the image appear to be within the -3.5 to 3 range.

It is likely there are 'no data' values in in Arc, and those pixels are ignored. Did you export the image from Arc at some point?

Check your mask - the white areas (where the pixels are all 1) should be the 'on' areas and black areas (where pixels are 0) will be masked. You can also try the ROI tool and create an ROI Threshold to see if you can locate the pixels outside of this range. The resulting ROI can also be used to create a mask.

Thanks, that solved the problem. I was able to make a new raster with the ROI as a mask, and check the statistics to make sure the image was valid. THOR worked as long as the image passed the statistic test.

For whatever reason, creating a threshold mask didn't work initially. It's probably user error, but I couldn't figure out how to invert the mask. So I could only mask areas between two values, instead of the areas outside the values. I also have an old version of envi.

There is a setting in the Build Mask tool where you can choose to have 'selected areas: On' and 'selected areas: Off'. This is how you can 'reverse' the mask.