M109 - Masking faint diffraction spikes?

[Aramcheck]

Hi,
I recently took 5 hours of subs on M109 using SW200dps & Canon 600d (astromodified) from our Bortle 7 back garden. When processing in Pixinsight I noticed what looked like an aircraft contrail running near vertically through the stacked image. Closer inspection shows that it is however a diffraction spike from Phecda (a Mag 2.4 star about 40 arcmins from M109).

Other than resorting to CloneStamp to paint over the area, is there any way to create a mask? I've tried taking a crude B&W mask on a cloned image & using Convolution to blur the edges, but the results still weren't perfect.

The spike can be seen more easily at the bottom half of the image.

Many thanks
Ivor

[ollypenrice]

'Clone Stamp' suggests you're a Photoshop user. Noel's Actions (now Pro-Digital Astronomy Tools, I think) has an action for removing vertical banding.  I would run the action on a bottom copy layer so that you could then go to the original top layer and erase the lines to reveal the repair only where it was needed. If you don't have the actions - you should! 

You could also make a copy layer and, again, work first on the bottom, very crudely clone stamping the lines and not bothering to avoid the stars. Just stamp them out. Then go to the top layer and use Colour Select to pick up the lines. This will pick up all sorts of other stuff as well but with the right sized eraser you could erase the top fairly quickly with no great need for accuracy.

Or yet again: copy layer and work on the bottom. Curves. Pin the curve at the background sky brightness and add a fixing point below that. Pull the curve down just above that till the lines have the same brightness as the background (ie they have disappeared.) Restore the curve above that to make the rest of the image fairly similar to the orignal. Go to top layer and erase the lines, maybe using the Colour Select tool as above.

Olly

[Aramcheck]

Thanks @ollypenrice. I had hoped to avoid a painting solution, as it goes against the Pixinsight ethos. (and my copy of PS is so old I have to run it on a virtual PC + it only works with 8 bit images.)

Your third option sounded the best for me, so I've just tried making a mask of the spike area using CloneStamp + Convolution to blur the edges. With the mask applied I then did a CurvesTransformation. That left the affected area slightly too dark, so I then used  Ron Brecher's PixelMath expression(s) for reducing DLSR mottle. That seems to have done the trick!

Many thanks
IVor

[ollypenrice]

2 hours ago, Aramcheck said:

Thanks @ollypenrice. I had hoped to avoid a painting solution, as it goes against the Pixinsight ethos. (and my copy of PS is so old I have to run it on a virtual PC + it only works with 8 bit images.)

Your third option sounded the best for me, so I've just tried making a mask of the spike area using CloneStamp + Convolution to blur the edges. With the mask applied I then did a CurvesTransformation. That left the affected area slightly too dark, so I then used  Ron Brecher's PixelMath expression(s) for reducing DLSR mottle. That seems to have done the trick!

Many thanks
IVor

Glad you've sorted it. I regard the 'Pixinsight ethos' as being pure tosh. It reminds me of those monks whose vows forbade them to handle money so they wore gloves. If the PI ethos is based on being 'true to the data' then your only option is to leave the spike in place. I see no difference between painting by numbers and painting with brushes! 

Olly

[CraigT82]

Once that data is stretched it ceases to be science, and becomes art, so after that you might as well use brushes!