New Tutorial: Deconvolution and HDRWaveletTransform with NGC 5189 Gemini South Image

[Juan Conejero]

Hi all,

A new tutorial about the Deconvolution and HDRWaveletTransform tools is now available (English only):

http://pixinsight.com/examples/deconvolution/Gemini-NGC5189/en.html

This tutorial makes use of new modules that will be released with the next version. For the impatient:

Linux 32-bit:
http://pixinsight.com/export/tmp/Deconvolution-pm32.so.tar.gz
http://pixinsight.com/export/tmp/MaskGeneration-pm32.so.tar.gz

Windows 32-bit:
http://pixinsight.com/export/tmp/Deconvolution-pm32.dll.zip
http://pixinsight.com/export/tmp/MaskGeneration-pm32.dll.zip

To install these modules, uncompress the archives and copy them to the bin installation directory. Then run Process > Modules > Install Modules, click Search, and then Install. This will install the new MaskGeneration module.

This tutorial is very large and includes detailed descriptions of many advanced techniques. It uses raw NGC 5189 data from the Gemini Science Archive.

Here's a direct link to the full-resolution processed image:
http://pixinsight.com/examples/deconvolution/Gemini-NGC5189/NGC5189.jpg

Enjoy!

[twade]

Juan,

Excellent tutorial!  I really like the new StarMask.  It is an excellent addition to PixInsight Standard.

When working with a monochrome h-alpha image, I've found that the Shadows, Midtones, and Highlights are the opposite of what you describe when creating a StarMask.  This makes perfect sense once it dawned on me as to what's going on.  

While creating a mask, I'm having difficulty protecting ALL the stars.  For example, after running through the deconvolution process, the small, unprotected stars get way too bright.  Is there a way to protect all the stars, and still have a good StarMask or is this just a casualty of the deconvolution process?  

Thanks,

Wade

[twade]

Juan,

I'm not sure if anybody else is having a problem with the tutorial loading completely on the latest version of Firefox, but it stops drawing text and graphics right after "Optimize Wavelet Regularization" section.  The graphic just below the paragraph beginning with "Wavelet regularization" is partially drawn and then Firefox stops drawing the text and graphics thereafter.  The remainder of the page is there, but it is blank.  Definitely some strange behavior.

Wade

[avastro]

Hi Twade & Juan,
I’ve the same behavior here using Firefox.

Antoine

[vicent_peris]

Hi all,

Juan has corrected the problem, you can now view entirely the document.

My congratulations to Juan on this excellent work!

And my greetings to the "Hubble's man" (just jokin' :lol: )


Best regards,
Vicent.

[David Serrano]

All fine here, with both seamonkey ("iceape" in Debian) and firefox ("iceweasel"):

Code: [Select]

- Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.8) Gecko/20071008 Iceape/1.1.5 (Debian-1.1.5-1)
- Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.8) Gecko/20071004 Iceweasel/2.0.0.8 (Debian-2.0.0.8-1)

[caliu]

¿Y para los que no sabemos ni papa de ingles? :cry:

[Juan Conejero]

Hi everybody,

Thank you all for your kind supporting words.

I've uploaded a new version that works correctly on Mozilla Firefox for Linux and Windows, on Konqueror (Linux), and on Microsoft Internet Explorer 7 (Windows).

The entire page is shown and all mouseovers work on all browsers. All index links (left side) jump also to the correct locations on all browsers.

Please let me know if you find more errors. Any corrections to my English are also welcome (remember that Spanish is my mother tongue).

[Juan Conejero]

¿Y para los que no sabemos ni papa de ingles? :cry:

Hola Caliu

De momento no puedo afrontar una traducción de este tutorial, que me costaría muchos días hacer. Pero prometo que la habrá lo antes posible.

Como te puedes imaginar, voy retrasadísimo con la documentación oficial de PI. Además hay todavía bugs pendientes y varias herramientas que tienen que salir y aún tengo que revisar. A todo esto hay que sumar la comercialización, que también lleva una cantidad de trabajo considerable. Todos mis esfuerzos están ahora concentrados en esto, pero como te decía prometo traducir este tutorial en cuanto pueda.

[Juan Conejero]

Excellent tutorial!  I really like the new StarMask.  It is an excellent addition to PixInsight Standard.

Thanks!

While creating a mask, I'm having difficulty protecting ALL the stars.  For example, after running through the deconvolution process, the small, unprotected stars get way too bright.  Is there a way to protect all the stars, and still have a good StarMask or is this just a casualty of the deconvolution process?

If I had to figure out a fire test for any star mask generation routine, one of your wide field images surely would be an excellent candidate

Deconvolution tries to unwise the effects of the PSF for the whole image. Since tiny stars are very close to true point sources as represented on the image, the end result is that all of them will be smaller and brighter after deconvolution. Indeed this may be an undesirable side effect, because too bright myriads of stars can steal protagonism from the main subject.

My suggestion is to try deconvolution with an inverse star mask active. Such a mask can also be built with StarMask. The idea is to protect stars from excessive sharpening. Note that this mask has nothing to do with the deringing support, which actually doesn't act as a mask (but as a working pattern for the deringing algorithm). In principle, I even don't see any problem in using the same star mask as both the deringing suppport for Deconvolution and as a mask to protect stars (with active mask inversion).

If you want, I could have a look at some of these images to suggest a suitable star mask.

[Jordi Gallego]

Hi,

excellent tutorial Juan, it will need some time to digest all information given! . Thanks and congratulations! :wink:

[My suggestion is to try deconvolution with an inverse star mask active. Such a mask can also be built with StarMask. The idea is to protect stars from excessive sharpening. Note that this mask has nothing to do with the deringing support, which actually doesn't act as a mask (but as a working pattern for the deringing algorithm). In principle, I even don't see any problem in using the same star mask as both the deringing suppport for Deconvolution and as a mask to protect stars (with active mask inversion).

This is very interesting, and we should give it a try, as problem pointed out by Wade seems to be also present in widefield DSLR images. Indeed the "starmask generator" is a great new tool

Regards
Jordi

[C. Sonnenstein]

It is a very complete and excellent tutorial Juan, I specially like so much how you have optimized the deringing support and wavelet regularization.

Thank you for you effort!!

[David Serrano]

About removing hotpixels... what about this?

- With PixelMath, create a mask that includes only hot pixels: $target[0] == 1 || $target[1] == 1 || $target[2] == 1 (this expression selects overexposed pixels too, but we could either live with that or refine it)
- Having the mask active, apply a convolution with the following 3x3 kernel: [ 1, 1, 1, 1, 0, 1, 1, 1, 1 ].

What I'm trying to do is substitute every hot pixel with an average of its surrounding pixels. How do you think this (untested) method would compare to using the CloneStamp tool?

--

Or, alternatively, in an oversampled image like that, why not just remove the first wavelet layer?

[OriolLehmkuhl]

Hi Juan,

We are new in this forum. These type of tutorials makes of PixInsight the best tool for the astronomical image processing  :wink:  Our Congratulations, and we will use the tutorial with our data, thanks!

Best regards,

Oriol and Ivette

[Maxi]

Gracias Juan, haber que me sale  :wink:

saludos