Deconvolution Rings at Star Centers

[themongoose85]

Anyone know why Deconvolution is causing bright rings to appear near the center of larger stars? I created a custom PSF using DynamicPSF, then created a star mask, and lastly a regular mask from an extracted L to only expose the stars and DSO. After I applied Deconvolution everything looked good under an STF stretch. After I did a MaskedStretch small rings appeared.

[Ignacio]

Unless you fully mask bright stars, you will always en up with some internal rings. If you only stretch using MaskedStretch, then those rings will persist. I typically do a regular, mild, MTF stretch, and then follow with a more aggressive MS, to avoid these rings. Alternatively, sometime I use the original but inverted starmask, to brighten those star cores and clear up the rings.

Hope it helps,
Ignacio

[themongoose85]

Thanks Ignacio. During an initial mild stretch with HistogramTransformation and following it up with a MaskedStretch seems to have resolved the problem.

[Carlos Milovic]

This donuts arise when you have clipped data, so the psf cannot model what is happening inside the cores of saturated stars. The plateu is considered as a real structure, and thus, edges are enhanced, creating the ring.

[themongoose85]

Is there a way to prevent this via settings in Decon Carlos and can you definie what you mean by clipped data? Thanks.

[Carlos Milovic]

No, i'm afraid that you must correct/prevent that outside of the deconvolution process. The easiest way is just to use a mask, to avoid modification of saturated (=clipped, in this case) data. You may create this mask by using the binarize process, and set it a bit below the highest pixel value.

[themongoose85]

Ok when do I use the Binarized mask during a MaskedStretch or when applying Decon?

[Carlos Milovic]

Deconvolution. With Masked Stretch you will have some other artifacts, mostly because the stars are clipped at a different point, depending on the filter transparency and exposure time. So, instead of a white core, you'll have a strong/weird coloration. I think that the best way to reduce this problems is simply to set the white point with the histogram transform to the point were all channels are clipped, not only one.
 Also you may use the RepairedHSVSeparation script, o even better, make an HDR capture setting, to avoid saturated stars.