Deconv deringing wrecks images ?

[neilfleming]

Hey there, guys.  I picked up a copy of PI the other day, and am just starting to scratch the surface.  I gave this deconvolution a try, and here is what I have so far.  I created a suitable star mask, then ran 3 iterations of Regularized Van Cittert.  I used "Deringing" settings of 0.2 for dark, and 0.05 for bright, with default settings for the rest.  (The star mask creation was the hard part.)

Before decon:  http://www.flemingastrophotography.com/astropics/misc/M81_L_PI_undeconned.tif
After decon: http://www.flemingastrophotography.com/astropics/misc/M81_L_PI_deconned.tif

Here is an animated GIF:  http://www.flemingastrophotography.com/astropics/misc/M81_L_PI_deconned.gif

A couple of questions:
1) For the Gaussian PSF, where are the settings determined?  Is it just guessing, or where/how is the PSF created?
2) Ditto for shape and aspect ratio?
3) After doing the decon on the linear data, I did a histogram stretch and saved the before and after as a 16-bit TIF so I could take into ImageReady to do the animated GIF.  The images looked posterized, as though they were 8-bit images.  I checked in Photoshop, and they had the same appearance.  What did I do wrong?

   Thanks...Neil

[Nocturnal]

Howdy Neil, glad you're giving PI a try.

[RBA]

Hi Neil,

AFAIK, Photoshop chokes on 32 bit float images saved in PixInsight, but I've never experienced any problems when saving as 16 bit int. 

Re Gaussian PSF, others will probably give you a better answer, but I think most people just use a blind deconvolution approach.
There's a script (Script -> Utilities -> Deconvolution Preview) that may be helpful. I usually just try different settings on a preview.

If you want to "read" the PSF from a star, you can use the External PSF method, using an image of a representative star as the "sample".
I've never really obtained very good results with it, but as you can tell, I'm not an expert in the subject.

This link may be helpful: http://blog.deepskycolors.com/PixInsight/Deconvolutions.html
It's based on v1.5 and the dialog box has changed a bit for v1.6 (mainly the Algorithm section), but hopefully it still makes sense for the most part.

Rogelio

[sleshin]

Hi Neil,

Glad to see you trying PI. Looks like you got very nice decon results though I couldn't see the artifacts you mentioned in the images you posted. I wonder if the problem relates to using the Reg Van Cittert algorithm rather than the Reg Richardson-Lucy. Did you try L-R and were the artifacts still present? I'm sure you saw Juan's very helpful mini tutorial on page 1 of this thread, has great suggestions for settings. If you haven't already, try making the star mask with just the default settings and then select that mask as the local support in the deringing section of the Decon tool. Also, try a much lower value for Global Dark, say 0.02 and keep Global bright at 0.0. Use Reg R-L with anywhere from 10 to 50 iterations.

And, welcome to the world of PI.

Steve

[RBA]

Local deringing is driven by a special image (usually a star mask) that is used to change the way deconvolution works at each iteration: the local deringing support. Local deringing support images are very easy to build: it usually suffices with a relatively rough star mask covering the brightest objects.

Señor Conejero... What is the difference between selecting a star mask as our local deringing image, and actually applying the mask directly to the image we're about to deconvolve?
I noticed they don't produce the same results.

[vicent_peris]

Local deringing is driven by a special image (usually a star mask) that is used to change the way deconvolution works at each iteration: the local deringing support. Local deringing support images are very easy to build: it usually suffices with a relatively rough star mask covering the brightest objects.

Señor Conejero... What is the difference between selecting a star mask as our local deringing image, and actually applying the mask directly to the image we're about to deconvolve?
I noticed they don't produce the same results.

Hi,

Local deringing builds a mask that is applied in each iteration. This is not the same as making 100 iterations and then mask the processed image. Because of the iterative nature of deconvolution, deringing degrades the image during the iterations. In fact, in the worst cases you can see how ringing is replicated as a kind of waves around some stars. So, with local deringing, the ringings are corrected in each iteration.


Regards,
Vicent.

[sreilly]

Now I'm confused. Apply the mask or only list it in the local support or both?

[RBA]

Local deringing builds a mask that is applied in each iteration. This is not the same as making 100 iterations and then mask the processed image. Because of the iterative nature of deconvolution, deringing degrades the image during the iterations. In fact, in the worst cases you can see how ringing is replicated as a kind of waves around some stars. So, with local deringing, the ringings are corrected in each iteration.

Got it. Makes sense. Yes, in each iteration is the key word.
Thanks!

[sreilly]

Just to see the difference I used the deconvolution method on the same image but the first was with the mask applied to the image by dragging the tab of the mask image over to the image, using the show mask button to hide the mask, listing the mask image under deringing local support and doing 50 iterations of RLR decon. Saved that image identified as mask applied. I then did a HST on the resulting image noting the values so I use the same on the next test image. I opened the same original image and mask and for decon I did the identical process except I did not apply the mask to the working image but did list it as before. HST was applied using the same settings as the first image. The two cropped images are displayed below. The one that had the mask applied and listed has a much smoother background while the image that only had the mask listed under local support has a molted background. This image is 340 minutes averaged combined luminance of NGC3628. The 3rd image is the original cropped.

My decon settings are in the screen shot below. Can someone explain what the proper sequence should be?

[sreilly]

Hi Neil,

AFAIK, Photoshop chokes on 32 bit float images saved in PixInsight, but I've never experienced any problems when saving as 16 bit int. 

Rogelio,

PS CS5 doesn't have this problem on my Win 7 Pro 64 computer. Not sure about earlier versions as I uninstall previous versions.

Steve

[sreilly]

I seem to remember seeing a post where you are able to take an small area of an image and apply a process to it to see how it will turn out saving time rather than applying it to the whole image. I can't seem to find that information or it doesn't exist! It could happen  Anyway I either dreamed it or misplaced it and either way doesn't help. Can someone point me in the proper direction please?

Thanks,
Steve

[sreilly]

Using the same settings as my previous post I did the same image this time wit 10 and then 20 iterations. The same hst settings on both and posted below. Maybe my data won't take higher iterations. Other than doing just this, is there a faster way to tell?

Thanks,

Steve

[pfile]

I seem to remember seeing a post where you are able to take an small area of an image and apply a process to it to see how it will turn out saving time rather than applying it to the whole image. I can't seem to find that information or it doesn't exist! It could happen  Anyway I either dreamed it or misplaced it and either way doesn't help. Can someone point me in the proper direction please?

Thanks,
Steve

you didn't dream it! just define a preview on some area of the main image. the preview will show up as a separate 'tab' along the left edge of the frame. click that tab and you'll be looking at just the preview; drag your decon triangle to the window when the preview is showing... et voila.

[sreilly]

you didn't dream it! just define a preview on some area of the main image. the preview will show up as a separate 'tab' along the left edge of the frame. click that tab and you'll be looking at just the preview; drag your decon triangle to the window when the preview is showing... et voila.

Is there a way to save a preview to test and compare each result or am I better off just creating a cropped image and using that to do fast tests?

Thanks,

[pfile]

you can just drag the preview's tab onto the desktop and that will create a new window with just the preview in it, so you can refer back to it. it won't have a mask though.

if you make a cropped image to work on, then you'll have to create a new mask that matches the crop, if you are doing something that requires a mask.