Starship Asterisk*

As nice as that picture is, it would be interesting to see a view from outside our galaxy by "removing" foreground stars. Maybe software could be written to identify them (sharp points of light) then mask them off.

daddyo wrote: ↑Mon Jun 18, 2018 7:58 pm As nice as that picture is, it would be interesting to see a view from outside our galaxy by "removing" foreground stars. Maybe software could be written to identify them (sharp points of light) then mask them off.

The trick isn't so much in removing them as it is in deciding what to replace them with--at least when working with a single photo. I've tried manually editing them out, but the hole that's left is just as much of a problem. Now if you could take two separate exposures from widely spaced locations, then you could subtract the bright spots that are slightly displaced when comparing one pic to the other. But I'm not sure photos taken even six months apart--at opposite sides of the Earth's orbit, in other words--would show sufficient movement of our local stars to allow that trick to work. Nor would it take into account the rotation angle of the spikes. It's a gnarly problem!

Rob

Hey Rob, I suppose you could take adjacent pixels that lie outside the detected star/cross occultation pattern and blur/average/interpolate them into the locations of the offending stars. Then it might just appear as slight smudges when inside the galaxy of interest, and black when outside. If needed some hand-editing of those regions by copying nearby patches might help. I can try coding something if interested. Seems like somebody would have tried this. Would be cool to have a "clean" view, even if we don't know what's behind those local stars.

I've long thought that at least the spikes--the most distracting parts, to me--could be algorithmically removed. Surely, in a field of dots that are effectively random in location, hue, and brightness, some way could be found to identify a set of radial straight lines and a bright spot in their centre. Maybe it's never been done because people like sparkly things?

Rob

rstevenson wrote: ↑Tue Jun 19, 2018 4:49 pm I've long thought that at least the spikes--the most distracting parts, to me--could be algorithmically removed. Surely, in a field of dots that are effectively random in location, hue, and brightness, some way could be found to identify a set of radial straight lines and a bright spot in their centre. Maybe it's never been done because people like sparkly things?

Imagers remove diffraction spikes all the time (and other imagers add them). It's a simple process that is largely automated by a number of available imaging tools. Of course, it's an aesthetic modification that slightly reduces the accuracy of the image.

Do they do it on the image, Chris? Or is it done, somehow, in the data prior to rendering the image? -- if that question makes any sense, since digital images are all data, all the time.

Rob

rstevenson wrote: ↑Wed Jun 20, 2018 1:51 pm Do they do it on the image, Chris? Or is it done, somehow, in the data prior to rendering the image? -- if that question makes any sense, since digital images are all data, all the time. :-)

I've never heard of anyone trying to do it using the raw data. It's usually one of the last steps in the aesthetic processing, either manually using some sort of cloning tool, or automatically with a plugin designed just for removing spikes.