Dear ERMAPPERS,
My student, Mauricio Baquero, and I have found what we think may be a bug
in the Version 6.2 ERMAPPER software that prevents us from correcting the
USGS' "error" of placing zeros instead of NULLs in the black border around
the image and saving the results in a corrected data set. The reason we
want NULLs in the black border is that we do not want the border pixels
included in the histograms and statistics of the frame. This throws off
the contrast enhancement stretches and any automatic algorithms that we may
have created to run on systematically corrected data, which all have border
pixels as NULLs (the ERMAPPER software will ignore NULLs, but not zeros).
To make this correction, we create a polygon that we call REGION0 that
approximately joins all four corners of the image data, then run a formula
that sets zeros to NULL outside of REGION0, but leaves zeros unchanged
inside the image (REGION0). The problem comes when we try to save this
result. When you Save As a new .ers dataset, you have a choice of leaving
the NULL Value subwindow on the "Save As ER Mapper Dataset" window as blank
(this changes all nulls to zeros) or placing some value (we will call it
NV) in the subwindow for the NULL Value, which makes the program select the
value of NV-256*I as the DN that gets changed to NULL, where I is the
highest integer possible that still keeps this value positive. For
instance, when you place 9999 in the subwindow, all DN values of 15
(because 9999-39*256=15) are changed to nulls. There is no program that I
know of that will Save As and keep the null values and other integers just
the same as they are in the image you are saving, without change. Adding
the 9999 to the subwindow is what the ERMAPPER user manual recommends for
that.
Could someone show us the right way to do this, or change the software to
be able to SAVE AS without changing the NULL value? Because of the USGS
"mistake" in creating images with zeros instead of NULLs in the boundaries,
this is an important and timely issue for all automatic algorithms that
must work on both types (terrain-corrected and systematically-corrected)
images of LANDSAT data.
Your Friend, Bob Vincent
-----------------------------------------------------------
To make changes to your subscription, please visit our website,
http://www.ermapper.com/technicl/ermapperl/index.htm
