Edward Kelly
New Member
Posts:6  
| 01 Sep 2017 08:55 AM |
|
|
Hi, I read somewhere in the academic literature that you can't use the MNF bands directly in the linear unmixing process and that you should use MNF inverse transformed bands as input? Is this true as I haven't seen it mentioned anywhere in the ENVI documentation.
Thanks.
|
|
|
|
MariM
Veteran Member
Posts:2396 
| 01 Sep 2017 10:16 AM |
|
|
This is true. When you are unmixing, you are trying to find the proportion of reflectivity that comes from a particular material within a pixel. So the point of converting to MNF space (and using the pixel purity index) is to assist the user in finding all the image endmembers to in LSU. However, you will use the endmembers with the original reflectance image to unmix.
|
|
|
|
Edward Kelly
New Member
Posts:6
| 01 Sep 2017 02:04 PM |
|
|
So I would use the Inverse MNF transform as input into the PPI and n-D Visualizer?
Thanks
|
|
|
|
MariM
Veteran Member
Posts:2396
| 01 Sep 2017 02:09 PM |
|
|
I would follow the Spectral Hourglass workflow - which assists you through the process to extract the endmembers. The endmembers are then applied in LSU to the original data. There is no need to do the inverse of the MNF data for this process.
|
|
|
|
Edward Kelly
New Member
Posts:6
| 01 Sep 2017 02:38 PM |
|
|
I was actually following this tutorial using ENVI classic, but it doesnt mention to use the Inverse MNF. I don't mind using the Inverse I am just unsure where to use it.
|
|
|
|