I would like to understand better the "Mask Count" variables under the ROI Statistics tool.

I am working with hyperspectral satellite images to which I didn't personally apply any mask. Nevertheless, when I design some ROI and ask ENVI 5.6 (or IDL 8.8 for what matters) to compute the ROI statistics, those are computed only on a subset of the ROI pixels.

For example I have a ROI with 11 pixels of which only 1 is used to compute the statistics (and, therefore, no standard deviation) and 10 are masked. The ROI Statistics panel, under "ROI Summary" reads "Pixel Count = 1" and "Mask Count = 10".

This behaviour disappears if I open the images with QGIS, save them back to ENVI format, and open them in ENVI/IDL again. In that case the ROI statistics are computed on the full amount of pixels in the ROI (i.e.: the ROI Summary reads "Pixel Count = 11" and there's no "Mask Count" variable anymore).

My guess was that some sort of bad pixels mask embedded in the image metadata and, if a given pixels is marked as bad in one of the bands, ENVI masks it. By manipulating the image through QGIS this metadata were purged and the masking disappeared alongside it. Nevertheless, if I use the Spectral Profile tool and move it around on all the pixels of the ROI, the spectrum always seems complete with no clear gaps in the plot.

Can somebody help me with this and explain such behaviour to me (or at least give me some hints on what to check to diagnose a potential problem)?

Thank you!