Ring Median methods


Index starters for this portion of the analysis:
Core, Sub, LPC, FCJ, 1, 3, 4, 7, Durrell


Again starting with Core region file


Using the IRAF "ring median" command RMEDIAN, the images were "smoothed."
This involves taking the median brightness around a pixel of a ring defined by its inner and outer radii, and replacing the pixel with that value.
Using 3 different inner and outer radius combinations, these sample images for Core were generated:

Examples:

/Core/CoreRmed1.fits,     Inner = 3, Outer = 5	/Core/CoreRmed2.fits,     Inner = 5, Outer = 9	/Core/CoreRmed3.fits,     Inner = 7, Outer = 11

(Again, clicking on any of these images will link to a page to track the editing process through the other regions)


From these trials, it was decided that the first file has too small an outer radius to be appropriate, and the second and third were used in continuing processing.


Next this smoothed image was subtracted from the original image resutling in an image with most of the galaxies (large-scale objects) removed

Example:
/Core/CoreRS2.fits
(full quality .fits files are included with the .jpgs on this page)


Next we need to remember that we already have 'bad' regions from previous methods (around bright stars/etc) and these should be masked out here as well.
These masks are used as the input Bad Pixel Mask (BPM) for the next step.

At this point there are two different routes to go. The simplest approach was just to set a cutoff for brightness on these Ring-Median-ed images, as (hopefully) only the stars would be bright enough to be masked out and the faint intra-cluster light would remain. This is detailed first. The second method involves OBJMASK again.




Masking by Brightness:


Index starters for this portion of the analysis:
Core, Sub, LPC, FCJ, 1, 3, 4, 7, Durrell

An easy next step to take with these Ring-Median-ed images is to create a mask based solely on the surface brightness, masking out any data above a certain level (i.e. things that are too bright to be galaxies: stars)

Using the equation    u = -2.5 log (I) + 29.2    (where 29.2 is a constant related to the imaging equipment/conditions), appropriate pixel intensities were calculated for u = 25 and u = 26.

Masks generated for Core: (non-black areas are "bad" and will be masked out on the original image)

/Core/CoreRS2u25.pl,      u = 25	/Core/CoreRS2u26.pl,      u = 26	/Core/CoreRS2u27.pl,      u = 27

(clicking on any of these images will link to a page that tracks this procedure through the other regions, u = 25, u = 26, u = 27)


Mask applied to Ring-Medianed (smoothed) image:

/Core/CoreRS2u25.pl on CoreRS2.fits	/Core/CoreRS2u26.pl on CoreRS2.fits	/Core/CoreRS2u27.pl on CoreRS2.fits

(clicking on these images will link to a page with the other regions' u = 27 masks. u = 25 and u = 26 are not included at this step, as u = 27 is preferable)


Mask applied to ORIGINAL region image:

/Core/CoreRS2u25.pl on Core.fits	/Core/CoreRS2u26.pl on Core.fits	/Core/CoreRS2u27.pl on Core.fits

Finally, just as a courtesy to the images, re-apply the manually made mask from before that blocks out the stars/rings/lines.

/Core/CoreRS2u25.pl and RMCore.pl on Core.fits	/Core/CoreRS2u26.pl and RMCore.pl on Core.fits	/Core/CoreRS2u27.pl and RMCore.pl on Core.fits

It is clear that this method does not even succeed at eliminating the stars. There are annoying halos around the medium-sized stars and the galaxies are basically un-touched.

A better masking method is.....






Masking by OBJMASK:


Index starters for this portion of the analysis:
Core, Sub, LPC, FCJ, 1, 3, 4, 7, Durrell


Alternatively, the same OBJMASKing methods were applied, but this time they are expected to work better as there are no difficult galaxies for OBJMASK to wrangle with.

Using all the same parameters as before (refer here)
As before there will be primary and secondary masks applied (P-mask and S-mask) with the P-mask to eliminate the bright stars and S-to eliminate fainter stars.

With blksize = -5 and hsigma = 2, the resulting mask is:
CoreRS2P1.pl (on CoreRS.fits)


At this point is is also useful to understand what this mask would look like on the original Core image, without the ring-median-subtraction:

CoreRS2P1.pl (on Core.fits)



When attacking it with a secondary mask next, we use a blksize = -20 and hsigma = 1, just as before and get:
CoreRS2S1.pl (on CoreRS.fits)


This secondary mask might appear to be overkill on the masking, but before throwing it out, it too, should be displayed over a normal (non-ring-median) Core region file.

CoreRS2S1.pl (on Core.fits)


Now it can be seen that this mask seems does an interesting job picking the right bits of the region to mask out.
Of note here is the Sub region's final Secondary mask. As there are huge galaxies dominating the frame, this method does not very well block out those major galaxies.




back to main Virgo page