Collecting Photons
Removing star halos
How to rescue haloes originating from large
stretching

Wide field image of the
cluster NGC7142 and of the reflections nebula NGC 7129 (VDB 146). The reflection
nebula itself is fairly rich in Herbig-Haro objects. The faint emission nebula
are the outskirt of SH2 137. I was very interested in depicting the subtle interplay of the dust
with the very faint emission nebula, that shows well in a wide field HLRG
composition, and this goal required very substantial stretching: this brought
about the problem that I will describe.
Astrograph ASA 10", SBIG sTL 11000.
Halfa 12x16';
Luminance 27x8';
R,B 12x10';
G 12x8'
for a total of over 12 hrs.
Processing of
this image has been very tricky, due to the large number of bright stars in the
field and to the amount of stretching that was required. Early on during processing, the application of the noise reduction
procedure of PixInsigh caused the formation of very faint haloes around the
brightest stars. Initially they were not visible, but, as stretching proceeded
they become apparent. Here I describe a procedure I devised to reduce their
impact. I am grateful to Paolo Lazzarini that pointed out the problem to me.
Back Home
|
1) A detail from the stretched luminance
Images were acquired with my ASA 10" astrograph
through Baader Luminance filter. Calibration, alignment, data rejection and sum
were performed in CCDstack and Registar. This image is the result of several
stages of stretching: the haloes, that were not visible at the time of the
initial noise reduction, are now quite apparent: rolling on the image with the
mouse will display an inverted and stretched image that will show the artefact
quite well.
|
|
2) Creating a mask for the selective
surgery
The general strategy I opted for can be summarized
in two points: 1) creation of a mask that selects for the haloes around the
bright stars. 2) application of the appropriate level corrections to the chroma
and luminance channels.
The decision of separating the processing on
luminance and chroma is quite important because it allows for a far wider
freedom of regulation. It also help in controlling star colour and saturation.
The first step in the generation of the mask,
consisted in extracting the star field from the original luminance. This is
performed by wavelet filtering removing all the spatial components larger than
11 pixels. This procedure leaves only the stars: if small residue of the
extended objects are still visible they can be removed in CS. Then this image is
clipped in order to suppress the weaker stars. The threshold is decided in a way
that only the stars surrounded by an halo are visible. This mask is expanded
with a dilation operator and low pass filtered. The parameters for these
operations must be decided based upon the size of the haloes. Finally, the core
of the star images must be protected by subtracting a properly scaled star mask.
On the left is shown this complex mask. |
|
3) Application of the corrections.
The correction levels
consists in histogram corrections applied through the above mask. I determined
that it was sufficient to modify the histogram midpoint to produce a satisfactory
halo cancellation. Obviously, some experimenting helps, here.
The screen
copy on the left shows the various levels applied in Photoshop CS.
The background
contains the RGB image. Its correction level is very subtle and helps in desaturating the residual halo artefact and in controlling the star colour.
Indeed, a second regulation is applied to desaturate a bit the haloes itself. On
top of the RGB layer sits the Luminance layer and its correction level that is
applied as a clipping mask (that means that the correction only applies on the
level immediately below). This regulation has a strong effect and must be
negotiated carefully.
|
|
4) Final result
The original image is shown on the left. If the mouse is rolled over, the
corrected image will now appear. It can be noted that the star haloes are far
less obvious and that the star themselves are less prominent, while the
background nebula is left unaltered. Final adjustments can now be applied to
this image. |
Send anything that comes to mind to Gimmi Ratto
gimmi@in.cnr.it
Copyright © 2009 by Gimmi Ratto. (August 31,
2009)