Welcome to the L3 Harris Geospatial documentation center. Here you will find reference guides and help documents.
﻿
>  Docs Center  >  Libraries  >  ASTROLIB  >  MEDARR

MEDARR

MEDARR

## Purpose

Compute the median at each pixel across a set of 2-d images

## Explanation

Each pixel in the output array contains the median of the
corresponding pixels in the input arrays. Useful, for example to
combine a stack of CCD images, while removing cosmic ray hits.
This routine became partially obsolete in V5.6 with the introduction
of the DIMENSION keyword to the intrinsic MEDIAN() function. However,
it is still useful if a input mask is needed (though it is much
faster to set invalid pixels to NaN values.)

## Inputs

inarr -- A three dimensional array containing the input arrays to
combine together. Each of the input arrays must be two
dimensional and must have the same dimensions. These arrays
should then be stacked together into a single 3-D array,
creating INARR.

## Optional Input

mask -- Same structure as inarr, byte array with 1b where
pixels are to be included, 0b where they are to be
excluded. For floating point images, it is much faster to
set masked pixels in inarr equal to !VALUES.F_NAN (see below),
rather than use the mask parameter.

## Outputs

outarr -- The output array. It will have dimensions equal to the
first two dimensions of the input array.
OPTIONAL OUPUT:
output_mask -- Same structure as outarr, byte array with 1b
pixels are valid, 0b where all the input pixels

## Restrictions

This procedure was *SLOW* when using the Mask parameter because it has
to loop over each pixel of the image.

## Example

Suppose one wants to combine three floating point 1024 x 1024 bias
frames which have been read into the IDL variables im1,im2,im3
IDL> bigim = fltarr(1024,1024,3) ;Create big array to hold images
IDL> bigim(0,0,0) = im1 & bigim(0,0,1) = im2 & bigim(0,0,2) = im2
IDL> medarr, bigim, avgbias
The variable avgbias will be the desired 1024x 1024 float image.

## Procedure

If the MASK parameter is not set, then MEDARR is just a wrapper for
MEDIAN(/EVEN, dimension = 3). If the MASK parameter is set,
a scalar median function over the third dimension is looped over
each pixel of the first two dimensions. The /EVEN keyword is used
with MEDIAN (which averages the two middle values), since this avoids
biasing the output for an even number of images.
Any values set to NAN (not a number) are ignored when computing the
median. If all values for a pixel location are NAN, then the median
is also returned as NAN.

## Modification History

Written by Michael R. Greason, STX, 12 June 1990.
Don't use MEDIAN function for even number of images.
W. Landsman Sep 1996
Converted to IDL V5.0 W. Landsman September 1997
Use /EVEN keyword to MEDIAN W. Landsman September 1997
Rearranged code for faster execution W. Landsman January 1998
Faster execution for odd number of images W. Landsman July 2000
V5.4 fix for change in SIZE() definition of undefined variable
W. Landsman/E. Young May 2001
Use MEDIAN(/DIMEN) for V5.6 or later W. Landsman November 2002