What version of ENVI are you using? In the latest versions, you will go to View Metadata->Edit Metadata. Here you will need to set the bad bands to 'bad' in the dialog. This requires you first determine which bands are bad by looking at the image spectra, statistics, etc. For example, if you see two bands that have the same mean and standard deviation, then it is likely it is a duplicate band. Or more commonly, if one of the bands is all 0 due to a bad sensor which is common in many hyperspectral sensors in the blue bands. You may need to set these bands to 'bad' in the header. Once you set the bad bands in the header, save it and try the process again.

Many thanks Mari!

I am using ENVI Classic 5.2 and I am preprocessing a Aster L1B image.

Basic Stats Min Max Mean Stdev
Band 1 0.026400 0.287000 0.098219 0.022197
Band 2 0.030100 0.352200 0.111724 0.027324
Band 3 0.039900 0.381000 0.122994 0.028320
Band 4 1.000000 1.000000 1.000000 0.000000
Band 5 0.890300 0.890300 0.890300 0.000000
Band 6 0.844000 0.844000 0.844000 0.000000
Band 7 0.921800 0.921900 0.921801 0.000012
Band 8 0.856500 0.856500 0.856500 0.000000
Band 9 1.000000 1.000000 1.000000 0.000000

I think bands 4, 5, 6, 8 and 9 are wrong.

I agree that those bands look incorrect. Are these the statistics from only opening the file in ENVI Classic? Did you pre-process the file in some way?

Yes, I did the following process:

1. Cross-Talk correction using SWIR Data IO Setup
2. Opening the file in ENVI Classic 5.2
3. Layerstaking VNIR-SWIR bands
4. Changing bands interleave
5. FLAASH atmospheric correction
6.Post-FLAASH Band Math
7. Resizing data
8. Reducing Noise (increasing S/N ratio)...... and here I get the problem.

What are the statistics of the radiance image you use for input to FLAASH and what are they after FLAASH? Somewhere along the way, the statistics return bad values. I am not sure what the Cross-talk correction is doing but it sounds like this result is used directly in FLAASH (after stacking and changing the interleave).

The statistics of the radiance image (input) are:

Basic Stats Min Max Mean Stdev
Band 1 0.000000 171.703995 54.467728 41.190968
Band 2 0.000000 179.832001 54.106559 42.184214
Band 3 0.000000 207.742004 41.847880 32.968428
Band 4 0.000000 55.219601 36.649617 26.087984
Band 5 0.000000 17.678400 11.729868 8.353174
Band 6 0.000000 15.875000 10.533317 7.501043
Band 7 0.000000 15.163800 10.064847 7.163811
Band 8 0.000000 10.591801 7.030525 5.003763
Band 9 0.000000 8.077201 5.361170 3.815902

And, the statistics of reflectance image (output) are:

Basic Stats Min Max Mean Stdev
Band 1 -858 4422 1079.866718 853.792764
Band 2 -82 5334 1404.605135 1120.454958
Band 3 -327 8662 1630.907828 1299.428456
Band 4 -98 10157 6696.156529 4777.412446
Band 5 -34 8909 5907.290458 4206.876722
Band 6 -30 8454 5590.705676 3991.352726
Band 7 -28 9238 6064.114314 4373.580403
Band 8 -24 8580 5648.903304 4058.845735
Band 9 -32 11872 7838.552624 5611.877334

Those look like they have valid ranges. Did you use a scale factor of 10 for FLAASH?
So after FLAASH, you performed Band Math, which would use an equation:
float(b1)/10000
to get to percent reflectance. How did you end up with all pixels having a value of 1.0 in band 4 and band 9? And all pixels in Band 6 having the exact same value 0.844?

Yes, I used scale factor of 10 for FLAASH and I used this equation post-FLAASH: (B1 le 0)*0+(B1 ge 10000)*1+(B1 gt 0 and B1 lt 10000)*float (b1)/10000.

I don´t have answers for your two finals questions. :(

I did the same pre-processing for another Aster image and it worked.

And what are the stats of using your band math equation? This is the likely place where the bands became all 1.s or the same value.

The stats post band math are:

Basic Stats Min Max Mean Stdev
Band 1 0.000000 0.442200 0.108000 0.085357
Band 2 0.000000 0.533400 0.140461 0.112045
Band 3 0.000000 0.866200 0.163095 0.129937
Band 4 0.000000 1.000000 0.662727 0.472779
Band 5 0.000000 0.890900 0.590735 0.420679
Band 6 0.000000 0.845400 0.559078 0.399125
Band 7 0.000000 0.923800 0.606425 0.437339
Band 8 0.000000 0.858000 0.564901 0.405870
Band 9 0.000000 1.000000 0.661149 0.473319

And this is the file you use as input to MNF and calculating noise statistics? Or is there another step? These are not the stats from the file you showed with the bands set to the same value.

Yes, one more step, and I think that here is the problem. I changed the size of image with Basic tools->Resize data and then I got the following stats:

Basic Stats Min Max Mean Stdev
Band 1 0.026400 0.287000 0.098219 0.022197
Band 2 0.030100 0.352200 0.111724 0.027324
Band 3 0.039900 0.381000 0.122994 0.028320
Band 4 1.000000 1.000000 1.000000 0.000000
Band 5 0.890300 0.890300 0.890300 0.000000
Band 6 0.844000 0.844000 0.844000 0.000000
Band 7 0.921800 0.921900 0.921801 0.000012
Band 8 0.856500 0.856500 0.856500 0.000000
Band 9 1.000000 1.000000 1.000000 0.000000

So you resized by dimensions/pixel size and some of the resulting bands are all 1s? I have not seen that behavior before. Did you use the Resize Data tool and what are the parameters did you use?

Yes, that is the situation. I used the Resize Data (Spatial/Spectral) tool, then I pushed "Spatial Subset" and then I tried selecting the area throught "map" and "file" and in both cases I got the same situation,

Hmm, it doesn't make any sense to me. Can you post the .hdr information from the file before resizing. I will create a dummy dataset and see if I can reproduce the issue.