Javascript Disabled Detected

You currently have javascript disabled. Several functions may not work. Please re-enable javascript to access full functionality.

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Processing Messier 33 in PixInsight - A Tutorial

Started by Madratter , Sep 08 2013 12:59 PM

Prev

Page 3 of 7
1
2
3
4
5

Next
»

This topic has been archived. This means that you cannot reply to this topic.

169 replies to this topic

#51 Jeff2011

Soyuz

Posts: 3,727
Joined: 01 Jan 2013

Posted 10 September 2013 - 10:33 AM

It does look like you have more control over the decon in PixInsight. At this stage in the processing, I don't think you can flip back and forth between StarTools since ST keeps track of the noise until the final noise reduction is done. If you flip back and forth, that noise tracking is lost, I think.

Back to top

#52 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 10:55 AM

GRB, there are a bunch of free tools. FitsLiberator gets mentioned fairly often, although I haven't used it.

FitsLibererator web page

StarTools is inexpensive and is a great value. It is only 60$ Australian.

It can be tried for free. You just cannot save the files.

StarTools web page

Although I primarily use PixInsight, I bought StarTools after I already had PixInsight, so that tells you I was pretty impressed with it. I liked some of the tools enough I thought it was a worthwhile addition to my tool chest.

DSS you already have and I have played with that. It is a good tool.

Jeff: It is no trouble going back and forth between PixInsight And StarTools at this stage. Once things are no longer linear, that is a different story.

You would simply save the PixInsight file as Tiff 16 bit with no compression. You then bring it into StarTools, do the deconvolution, and save as Tiff. You do lose some precision in the process because you have down converted from 32 bit floating point to 16 bit integer. But still it works well.

What IS true is that once you have stretched the data permanently (not just a Screen Transfer Function or [ctrl]a), StarTools will not be able to optimize its noise reduction. And processes like deconvolution that depend on linear data to work correctly will not work as well. You could still try importing the data and doing a reverse stretch which StarTools will try.

Back to top

#53 Jeff2011

Soyuz

Posts: 3,727
Joined: 01 Jan 2013

Posted 10 September 2013 - 11:06 AM

Hmmm. Perhaps my understanding is wrong. I thought that ST kept track of the noise after each operation while still in a linear state and that noise tracking was used when tracking was ended and the denoise was applied taking the image out of a linear state. Perhaps Ivo can clarify this. To me any operation done in PI can't be tracked by ST. Plus if you do some operations in ST and then leave ST, I think the noise tracking is lost for any operations done so far. I thought I saw Ivo mention this before either on his support site or a thread on CN. Again I could be completely wrong here. Also I am using his latest production version and not the Beta.

Back to top

#54 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 11:12 AM

What you are saying is technically correct. That said, if you were to transfer over to StarTools and then continue on in StarTools after having done the cropping, and DBE in PixInsight, StarTools would still be able to track additional changes and would still optimize noise reduction better than if it got the image in a non-linear state.

BTW, the key word here is optimize. Even if you bring over a non-linear image, the noise reduction in StarTools will still work. In fact, I have done it.

Back to top

#55 Jeff2011

Soyuz

Posts: 3,727
Joined: 01 Jan 2013

Posted 10 September 2013 - 11:27 AM

Ok, that makes sense. To take full advantage of the noise tracking in ST you would do all of the PI operations first and then finish in ST. By finish I mean take it to a non-linear state. I realize there can be more processing after that in either PI or ST.

In ST I usually do the initial noise reduction that ends tracking on the whole image with nothing masked off. Then if the background is still noisy, I do another noise reduction (now in a non-linear state) with the DSO masked off, so I can do a more aggressive noise reduction on the background.

This has been a great discussion. Thanks.

Back to top

#56 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 08:57 PM

If you have saved your 4 files needed for deconvolution and are coming back to this like I am, it is time to load them again. You need the following:

1) The image of M33 we have been working on
2) The PSF image.
3) The star mask
4) The mask for protecting the background (deconv_mask)

Now you want to select the little tab that has the name of your deconv_mask (whatever you called it) and holding down the left mouse button, you want to drag it over to just below the tab with the name of your main image (in my case L_dbe). Then release the mouse button. This should apply the mask to your main image.

Attached Thumbnails

Back to top

#57 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:03 PM

Alternatively, right click on your main image. This should bring up a menu. One of the items is Mask and it has a submenu. From there you can "Select Mask". Then you will get a little window that will let you choose from your various images of the right size for your mask. If you happen to love shortcuts, you can bypass all of that and get right to the mask selection window with [ctrl]m. Personally, I use the first method of dragging the little tab.

If you have done this correctly, your main image will now have a red overlay that gives you an idea of what is masked and what isn't.

Attached Thumbnails

Back to top

#58 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:09 PM

That is great to know one is applied, but fortunately, we can turn off its display. To do this you can either use the mask menu or you can just use [ctrl]k. I use the later.

You can tell a mask is still applied because the tab of the image name will still be colored brown.

Now bring up the deconvolution process. Click on the tab saying external PSF. Then click the little box next to view identifier and select your PSF image.

Attached Thumbnails

Back to top

#59 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:12 PM

Finally, expand the deringing settings and click on the checkbox to enable deringing. Set Global Dark to 0.0. Check the local deringing box. Click the box next to Local Support and select your star mask.

Attached Thumbnails

Back to top

#60 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:15 PM

Now under the Algorithm section, change the iterations to something more useful like 50. This is one parameter you might want to play with.

Attached Thumbnails

Back to top

#61 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:21 PM

Next we need to define a preview. We do this exactly like we did before ([alt]n left click and drag mouse to define opposite corners). Here is my preview area. You definitely want to preview this because these calculations take a long time on the full image and we may need to adjust deringing to get something reasonable.

Remember to do a screen transfer function on the main image so we can see what we are doing.

Here is my preview defined. That preview area is a good one because:

1) It has some really bright stars (which are easy to mess up)
2) Some regular stars.
3) Detail in the galaxy to be enhanced.
4) Background we don't want messed up.

Attached Thumbnails

Back to top

#62 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:25 PM

Select the Preview tab and we are ready to actually try this out. You might want to expand the window some if your preview area is as large as mine so you can see all 4 of those features.

Attached Thumbnails

Back to top

#63 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:29 PM

Now drag the little triangle at the bottom left of the deconvolution window over your preview and let the calculations run. My result is this:

Attached Thumbnails

Back to top

#64 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:33 PM

This is already a much better result than when we tried deconvolution with the default settings. Star sizes have shunk nicely, detail in the galaxy is enhanced (especially in that HA region), the background has been protected, and the big stars don't have big holes or anything.

You can toggle back and forth looking at the original and new versions using [ctrl][shift]z.

Back to top

#65 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:38 PM

That said, I'm not entirely happy with it. I tried setting the global dark reringing parameter to .01 and I like the result much better.

Attached Thumbnails

Back to top

#66 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:47 PM

I'm now happy with what I have and I'm going to apply these settings to the image as a whole. I select the tab for the regular image (L_dbe in my case) and then drag the little triangle at the bottom left of deconvolution to my image. This will take a longer time to calculate. It should now be obvious why I used a preview in making my adjustments. It takes a couple minutes to calculate the whole image on my computer.

Attached Thumbnails

Back to top

#67 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:52 PM

As soon as it finishes, I save my image (I used L_deconv) as a 32 bit fit.

You can compare the before and after by using the Undo and Redo buttons at the top right of PixInsight.

You might want to zoom in some using your middle mouse wheel when doing these comparisons. Don't forget to make sure you Redo as your last step. Otherwise, you might continue processing on the image before you did the deconvolution and that would waste all this work.

Attached Thumbnails

Back to top

#68 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 10 September 2013 - 09:55 PM

Next up, we are finally ready to stretch our image. But this is a good stopping point for this evening. Besides I'm collecting data on another project tonight. Automation software is wonderful.

Back to top

#69 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 12 September 2013 - 06:31 PM

It is now time to stretch the luminosity data. Bring up the HistogramTransform process. You want to blue check mark on the bottom right. That enables track view. What it buys you in practice is that changes you make will now be displayed in the upper histogram (i.e. what the histogram will now look like). You also want to press the blue circle in the bottom left of the window. That will bring up a preview window. Resize that window so you can see what you are doing.

Attached Thumbnails

Back to top

#70 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 12 September 2013 - 06:41 PM

Make sure you have reset the ScreenTransferFunction. Now move the middle slider for the Histogram Transformation over towards the left. You'll want to be quite aggressive with it.

Attached Thumbnails

Back to top

#71 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 12 September 2013 - 06:42 PM

Here is what the real-time preview looks like.

Attached Thumbnails

Back to top

#72 Madratter

Hubble

topic starter

Posts: 13,277
Joined: 14 Jan 2013

Posted 12 September 2013 - 06:49 PM

Now close the real-time preview and drag the blue triangle on the bottom left of the HistogramTranformation window over the image. Alternatively, click the blue box on bottom left of the HistogramTransformation window to apply it. My image now looks like this.