Using Dynamic PSF to add some clarity

[NickK]

M57.. @ 1370mm fl (at 0.88 arcsec/pix which is below 1.14 arcsec/pix dawes limit of the scope).. 8 images without darks or flats. I've done my form of drizzle - 4.X & 4.y upsize, 45 degree rotate, registration and then stacking..

Select each star by simply clicking on each star as a source for each PSF:

Then use this button to create an average artificial PSF from the stars you've selected previously:

We're given the generated PSF as a new image which is very useful

Close the Dynamic PSF window by using the red cross at the bottom - this will clear the lock against the window preventing the image from being altered.

Now take the deconvolution mechanism, select the "External PSF" tab and then select the generated PSF image (here called PSF1):

Then apply by dragging the triangle on the bottom left

By using undo and reapplying using different values you can play around..

Now we can use the Dynamic Range Extension too...

Not bad reclamation of detail from the image.

[NickK]

I forgot to add the images (at native resolution) - click for actual size...

Before:

After:

[NickK]

Using the motion blur PSF to reduce the elongation of the stars and reapplying a second Dynamic PSF again:

[squeaky]

Hmmm....

Not bad.

You have a difficult image to work with there - in that your brightest star is somewhat oval, the middle range stars look more like distant fuzzy galaxies, yet the smaller fainter stars appear to be round.

Anyway - here's a slightly different technique (in PS) based on star shaping that applies to the whole image. This has the added benefit of redefining the nebula as well, and it also tends to reduce background noise. Once done, three iteration of a high pass sharpen, one wide (65px) then two smaller (6px and 3px). I should probably have had a go at levels and curves before even starting these things but I just wanted a quick play.

[NickK]

There's a modification to the technique I'd like to do - closely resembling the guider-based DPSF where I'd DPSF each sub after the upsize and the rotation.

There's definitely something to continuous PSF deconvolution...

[squeaky]

Aye. There's loads of good data there, it's just a case of finding the best way to bring it out. But that really is a difficult image to work with given the different ways that different brightness stars have displayed differently. I didn't really find a way to sort them all out, and although picking them individually might be OK for an image like this one with not many stars - try that technique on M13 and you'll be busy for months

Applying your routine, or some part of it, on the subs before stacking might well be the way to go? Lot of work though. It's going to be interesting to see the final result. For a bet you'll finish up with a darn good show. Not there's owt wrong with what you have

[NickK]

Just to add - if you select all the stars in the list then create the Dynamic PSF then you'll get an average. Selecting specific stars without any noise and getting an average results in a very good result.

[Helen]

Hmmm....

Not bad.

You have a difficult image to work with there - in that your brightest star is somewhat oval, the middle range stars look more like distant fuzzy galaxies, yet the smaller fainter stars appear to be round.

Anyway - here's a slightly different technique (in PS) based on star shaping that applies to the whole image. This has the added benefit of redefining the nebula as well, and it also tends to reduce background noise. Once done, three iteration of a high pass sharpen, one wide (65px) then two smaller (6px and 3px). I should probably have had a go at levels and curves before even starting these things but I just wanted a quick play.

Would you mind explaining a bit more about your 'different technique in PS'? It looks very powerful!

Thanks

Helen

[squeaky]

Would you mind explaining a bit more about your 'different technique in PS'? It looks very powerful!

Thanks

Helen

Hiya

I don't really want to take over this thread so I'll start a new one. Right after I've grabbed some screen shots.

EDIT: Just had a quick try and wasn't able to duplicate my result. I can't remember everything I did. I'm going to need to spend some time on it again and work it all out. There's no set "standard" approach - no specific numbers on settings to aim at - I have to "feel for it" to deal with whatever the image has to offer. This one was really difficult. A general technique gets a fair result, but I'd like to do a step by step to get back to exactly what I got in my example. In the meantime... I've got my dob mount in pieces all over the kitchen floor, so this will have to wait for another day. Please don't hesitate to give me a nudge by PM in a couple of days if I haven't caught up with it

[Helen]

Thank you - no rush

Helen

[NickK]

Thank you - no rush

Helen

Found out today in WHSmiths there's an article in S@N magazine that goes through deconvolution. It seemed to concentrate on 'straight' deconvolution which uses an ideal maths function (a lot to be desired in my view).

Maybe I should write a "The Point Spread Function and what did it ever do for us?" article..

[squeaky]

Found out today in WHSmiths there's an article in S@N magazine that goes through deconvolution. It seemed to concentrate on 'straight' deconvolution which uses an ideal maths function (a lot to be desired in my view).

Maybe I should write a "The Point Spread Function and what did it ever do for us?" article..

But you're only allowed ONE equation

[Helen]

But you're only allowed ONE equation

I like equations

Helen

[squeaky]

I like equations

Helen

I was referring back to Hawkins' "A Brief History of Time" where he was only allowed the one If he can do it when writing about the beginning of the Universe....

[NickK]

I was referring back to Hawkins' "A Brief History of Time" where he was only allowed the one If he can do it when writing about the beginning of the Universe....

Hawkins you say...

image = f( target object )

are we allowed vector spaces?

Seriously it's not that bad and means more without referring to maths and what you're attempting todo..

[squeaky]

Hmm...

"E=Mc^5... nahhh... E=Mc^4... nahh... E=Mc^3..... ah, blow it."

[NickK]

Hmm...

"E=Mc^5... nahhh... E=Mc^4... nahh... E=Mc^3..... ah, blow it."

Hehe, I think you know that the real definition is constantly changing and any definition is an approximation. (entire universe over time)

[squeaky]

The only absolute that I'm (reasonably) sure of is absolute zero. Not necessarily the number of deg K that is quoted - but that there is one.

Do you know where the coldest place in the entire Universe is? I do

[Helen]

Do you know where the coldest place in the entire Universe is? I do

My husband would probably argue it was my feet after a night observing in winter

Helen

[squeaky]

You're a whole lot closer to the right answer than you'd think

[NickK]

I've just found this very nice (and more advanced method): http://www.manuelj.c...53866&k=NKhtp5J

Basically this uses a mask, then uses the stars from the mask as the source of the D-PSF. It also puts together a nice work flow. There's also a point about reducing the standard deviation and increasing the number of iterations to reduce the black halos that occur.

I'm tempted to add the motion blur deconvolution to this - showing the use of the D-PSF parameters as the input to the motion blur - and write a full on deconvolution guide.