X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 11 Jul 2005 05:49 AM by  anon
claculation of true probabilities from maximum likelihood rule images
 2 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
11 Jul 2005 05:49 AM
    How I can generate images of true probabilities from rule images of maximum likelihood classification, directly (or programmically) and not by RULE-CLASSIFIER module of ENVI?

    Deleted User



    New Member


    Posts:
    New Member


    --
    11 Jul 2005 05:49 AM
    Sorry, the following is more an answer than a response: I understand that each rule image (one by class) shows the probability of pixel belonging to class. So we can expect a value between 0 and 1 and the sum of all the probabilities for each pixel will be one. However when I look at the values of the pixels in rule image (produced by maximum likelihodd) I get negative values. I saw that you posted your question 2 months ago, about max. likelihood rule images. Do you know some way to interpret rule images ? Best Regards, JF

    Deleted User



    New Member


    Posts:
    New Member


    --
    11 Jul 2005 05:49 AM
    sorry, in my last message I was wrong. For the Maximum Likelihood Classification, ENVI calculates the discriminant functions for each pixel gi(x). The probability is gi(x)/(sum over i of all gi(x)). So it seems that using tools as "math band" it is possible to calculate the probabilities (or exporting to some other package such as grid in arc/info to carry the computing of the probabilities. There is more detauils in http://www.rsinc.com/serv...asp?ttid=1793&pv=yes
    You are not authorized to post a reply.