There is nothing in ENVI that I know of. But you might find some interesting info by googling "Shadow Insensitive Classification". I found some articles about approaches using hyperspectral imagery at the top of that search. I didn't read closely, but it sounds like some of them might also apply to multispectral. I think you're going to have trouble with this if you don't have spectral imagery, though. It occurs to me that if you have the EX module, the Feature Extraction workflow might be something you could use to at least identify the mountain shadows (as opposed to other dark features). Not sure what you could do for the next step, though, to actually get good classifications inside the shadowed areas. There's not much signal from deep mountain shadows.

Hi, There is an ENVI pro filed called C-Correction that you can download that supposedly corrects for topographic variations using A DEM but I can't get it to work! Cheers Jeff H

In hopes that this helps the next person to come to this post. I believe the above post refers to C-Correction code written by Mort Canty for his book: Image Analysis, Classification and Change Detection in Remote Sensing, with Algorithms for ENVI/IDL and Python, Third revised edition. The code can be found here: http://mortcanty.github.io/src/software.html The C-Correction code (found in Chapter 5) does work, but will have errors and output NaN values if the DEM isn't larger than or exactly the same size as the image (or subset) you input into the algorithm. One way to avoid this is to make sure the DEM Raster is larger than your study area (the algorithm will automatically subset the DEM). Another solution is to subset your image based on the DEM coverage when you select your input image. I've run the C-Correction code with ENVI 5.2 and can verify that it does work.