White circular patch on the stars

[Bobinius]

Hello everyone,

I observed an artefact during post-processing which I believe it is due to an imperfect mask during deconvolution. The stars have in their center (visible on the big  ones) a solid white circular area which is homogenous and saturated white? Is it a known effect of some imperfect setting? How could I improve that ?

Thanx.

B

[pfile]

you might try a local support image for Deconvolution. this problem is common around stars. or, maybe make your mask stronger so the stars are not affected as much by the deconvolution process.

rob

[ngc1535]

If I interpret the OP correctly- the issue is the local support image. Typically a mask directs what bright signal is going to have deconvolution applied (including bright stars)... not the protection of stars (at least this is how I approach things). However the local support image can be used to moderate the effect the OP describes on nearly saturated (or non-linear) stars (basically stars with non-conforming PSFs). So a "stronger" local support image (meaning brighter stars in the image) will help prevent the brightest stars from having an unnaturally concentrated core. A combination of brighter stars int he local support image...and perhaps even convolving them to be larger will help quite a bit!  Local support is not a mask... so if I am interpreting things correctly, this might be helpful.
-adam

[Bobinius]

Thanx for your answers. Yes I used a local support image or a mask that is not a mask... I don't really understand the details of how this image interacts with the deconvolution process since it is not a mask (and it does not protect). So should I use a transformation of the support image with curves of Histogram transformation to make it more saturated ? How to you transform or adapt your local support images?

[wadeh237]

For a deconvolution support mask, I start with the linear image that I am going to deconvolve.

I use a RangeSelection to capture the very brightest stars in the image.  I typically do a real time preview as I adjust it to make sure that it picks up the stars that I am interested in.  In the case of a galaxy with a really bright core, I have seen the range selection pick up the core.  If that happens, I just clone stamp it out.

From the mask generated by the RangeSelection, I then do a MorphologicalTransformation with dilation to enlarge the points so that they are big enough to cover my bright stars.  I finally do a Convolution operation to blur the edges.  For my particular image scale and image data, I find that doing MorphologicalTransformation with size 25 and a circular kernel works pretty well.  And I do Convolution StdDev 2 and Shape 2 for the blurring.

The resulting local support image does pretty much exactly what I want it to do during Deconvolution.

[Bobinius]

Ok, I used a star mask and enlarged the structures in order to capture the big stars. The problem is that since it is a Head Horse Nebula Alnitak is very big and not detected as a star. However, the saturated circular core patch is seen on other stars. I'll try to adjust with the range selection see how it goes.

Will deconvolution have a tendency to saturate the core? Ringing is pretty obvious as a side effect since it tries to improve the PSF and narrow it but does that increase the saturation of the core?

[pfile]

yes it tends to saturate the cores. what deconvolution is doing is concentrating the energy of the star more toward the center, since in essence it is trying to undo the blurring in in the image.

rob

[Bobinius]

Thanx Rob. So what is the solution for preventing that? I mean, what are the characteristics of a good deringing support image?

[Bobinius]

Or is it the Global bright parameter in Deconvolution/Deringing ? I understand it as a protection for the peripheral bright artefacts around the stars not at their core.

[wadeh237]

For me, using a support image as I described above *completely* resolves the issue that you are describing.

To the nearest that I can tell, the local support image protects the areas covered from any deconvolution (or at least significantly reduces it).  For really bright stars, this seems like the best thing to do.  If you want to make them smaller, the you could use MorphologicalTransformation late in processing steps after the image has been stretched.  It's usually one of the last things that I do (if I'm going to do it at all).

Deconvolution works great on dimmer stars and contrast details in non-stellar structures, but my personal opinion is that I think that overly bright stars should be left alone - and the local support image works great for that, especially if the bright star is embedded in galaxy or nebula details.

YMMV.

[pfile]

if the local support image is not working to your satisfaction you can also mask the cores of the stars with a regular old starmask...

rob

[Bobinius]

Ok, thanx guys, I'll let you if I manage to avoid this effect by modifying the masks.