[daip] misbehavior of "mode" filter in MWFLT??

[Susan Neff]

Hi Eric,


I think I have found a "funny" in MWFLT.  I was running it as follows:

opty 'mode'
cparm xx, xx, 0  (range from 3 to 61)
dparm 0, 1, 0


I believe this means that it should perform a median-weight filter, using
a box of size xx, xx, but using the mode instead of the median of the
surrounding points, and that it should write out the filtered image.  i.e.
it should give me an image that is somewhat smoother than the input
image - a "sky" image if the box is big enough with point sources 
suppressed.
Each pixel in the filtered image has been replaced by the mode (median if
opcode is 'mw')     of the pixels (in the original image)      that are 
in the specified box.


MWFLT is behaving as if it were doing some sort of subtraction in this
case - perhaps an offset subtraction or a row filtering or something.  
Examples
of this strange result are in user number 248, disk 1, and are called  
CEN2L_MWFLT.MODE.XX    where XX is the boxsize. 


When I run the task with opty 'mw' (for median instead of mode) it does
not seem to do this subtraction - examples are in the same place, but called
CEN2L_MWFLT.MEDN.XX


The original file for both of these is  CEN_2L_SUB.N14700.1400,
same disk.


Perhaps I am misunderstanding what a median filter does??  There is
something in the explain file about data +/- filter - I am interpreting
the +/- as what is controlled by the DPARMS.


Any help appreciated, as always,
-s-