X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 02 Aug 2006 07:56 PM by  anon
texture measure
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
02 Aug 2006 07:56 PM
    I am trying to use the occurence and coocurrence statistics calculated by envi (texture filters). As far as I can tell, these methods use the nodata value as if it were simply a valid value. This seems wrong to me. (For example, any window with just one pixel of nodata value=-32767 will have its mean and all other statistics heavily influenced by such a value.) Can anyone verify this? I'd love any references to why this is the correct thing to do, or alternate strategies for calculating these stats while handling nodata pixels. Thank you kindly. -Tracey Marsh

    Deleted User



    New Member


    Posts:
    New Member


    --
    02 Aug 2006 07:56 PM
    ENVI uses the "no data" value in its calculations because the processing involved is kernel-based. To mitigate the effects of the "no data" value, you can try two approaches: 1) Build and apply a mask that will sets these values to something more reasonable. ENVI's masking tools can be found under the Basic Tools menu group. 2) Use the "Replace Bad Values" tool in the Topographic menu group to fill in these values with reasonable approximations based on surrounding "good" data. This is especially useful if you are working with data from missions like SRTM that tend to have voids that are arbitrarily filled with values of -32767.
    You are not authorized to post a reply.