X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 26 Apr 2017 02:12 PM by  MariM
Intrepretation of MNF eigenvector matrix
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

Rami Piiroinen



New Member


Posts:1
New Member


--
26 Apr 2017 09:48 AM
    Hi all,

    I performed minimum noise fraction transformation on a hyperspectral image with 129 bands. This produces 129 new bands, as it should. Okay, then I get the eigenvalues that I can use to interpret which bands contain information and which bands contain noise.

    Now I need to understand which bands constituted most to each new MNF band. To my understanding we need to look at the eigenvector matrix to see the "loading" from each original band. But, which way should we look at the matrix. Is each column representing new MNF band and rows the original bands, or vice versa? Essentially, how should I look at the eigenvector matrix, so that I can understand which bands constituted the most to each new MNF component.

    Br,
    Rami


    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    26 Apr 2017 02:12 PM
    Although the MNF eigenvectors are not exactly the same as PCA eigenvectors due to the noise whitening that is performed with MNF, I believe you can still consider the eigenvectors as the contributions to each PCA band.
    This help article discusses the MNF statistics files, which are a little different than PCA:

    http://www.harrisgeospati...D/19433/Default.aspx

    And this help article discusses how to convert PCA eigenvectors to "weights''::

    http://www.harrisgeospati...D/19426/Default.aspx

    The eigenvector matrix will contain the eigenvectors as the rows with the columns representing the band contribution for each vector.
    You are not authorized to post a reply.