Large Scale Wavelet Processing

[Radar]

Hi P.Team

I want to process an image to bring out a large structure in the Eta Carina Nebula. http://www.naturespeak.com.au/Eta%20Carina%20Nebula.htm

I want to be able to do large scale wavelet processing for this image. Basically what is in this example is what I want to do.  

http://www.pleiades-astrophoto.com/examples/B142-vperis/en.html

I can generate previews and assign a wavelet for each preview.

The wavelet layer that seems to hold the larger image scale is number four.   So I want to modify that layer (layer four) then place it back in the original image which will hopefully bring out Eta Carina's tendrils better.  

I am not sure how to do this.

Could you point me in the right direction please?
 
Thanks

Ray

[Juan Conejero]

Hi Ray,

Take a look at this excellent tutorial by Vicent Peris, our multiscale processing guru:

http://pleiades-astrophoto.com/tutorials/LE/M33-vperis/en.html

We have uploaded this tutorial just a few minutes ago. I hope it will help you to understand basic concepts of multiscale processing applied to deep-sky images. Should you have further questions, please post them here.

Cheers,
Juan

[Apollo]

What a great tutorial Vicent! Extremely well written and it's nice to see how to match the tools of PixInsight to practical examples. I think i'll be re-processing some of my images using a bunch of these techniques

Maybe you could explain how you came up with your initial chromatic balance adjustment values - it's something that I tend to find quite tricky. I guess it's because you're trying to juggle the 3 channel parameters at once.

I was also wondering if it is better to combine RGB channels then extract a Luminance channel, or rather sum or average the filtered views to create your "Luminance" channel? I suspect it does not actually matter if you have a good chromatic balanced before using either technique, although averaging rather than summing will produce slightly different (not so good) results. Has anyone got any thought on this?

Simon.

[Radar]

Hey Juan,

Thanks for the tutorial.

What are POSS plates?

Also, can this tutorial be used on a normal rgb jpeg image?

Cheers

Ray

[Radar]

I know this is a silly question, but I can't find wavelets on the Beta Version of Pixinsight 1.0.

Am I using the right version?

Thanks

Ray

[Juan Conejero]

Hi Ray,

What are POSS plates?

POSS = Palomar Observatory Sky Survey.

The POSS is a fundamental reference work, consisting in a vast collection of photographic plates taken with large Schmidt telescopes, covering most of the sky.

Here is a page that explains it in detail:

http://www.astro.caltech.edu/~wws/poss2.html

Very happily, the POSS II has been digitized and is freely available on the Internet. This is the interface:

http://archive.stsci.edu/cgi-bin/dss_form

On the page above, you enter the coordinates of the center of a region of interest (or you can enter the name of an object, e.g. M33, and get its coordinates from the SIMBAD database), then select a channel and a size in arcminutes (1 square degree maximum), and you can retrieve a POSS image in FITS format. Fantastic, isn't it?

The recommended channels are:
Red: POSS2/UKSTU R
Blue: POSS2/UKSTU B
Green: Quick V

Unfortunately, Quick V plates were taken with a (comparatively) much smaller focal length and in much poorer conditions. There are also some regions that lack the V channel at all. Then one has to synthesize a green channel from red and blue.

The digitized POSS is a great source of high-quality raw data for processing. We use it all the time to test countless processing algorithms and techniques.

Also, can this tutorial be used on a normal rgb jpeg image?

Sure, this technique can be applied to virtually any deep-sky image. However, jpeg images usually include compression artifacts that can be very disturbing. It's better to work with original tiff or fits images.

Cheers,
Juan

[Juan Conejero]

I can't find wavelets on the Beta Version of Pixinsight 1.0.

We still have not implemented a wavelets process on the new PixInsight Standard platform. We'll do so quite soon, however, along with a lot of other very interesting things, including deconvolution.

So for now, keep your LE application. Of course you can have both applications (LE + Std) installed. They don't interact mutually at all.

Regards,
Juan

[Radar]

Hi Juan,

Thanks for your reply.

Yes, that Sky Survey is quite impressive.

Would it be possible to get the same images of M33 that Vicent used in the Multiscale Wavelet Tutorial? When I try to download M33 from the Sky Survey my computer can't recognise the format.

http://www.pleiades-astrophoto.com/tutorials/LE/M33-vperis/en.html

Having the same images available for the tutorial would help. I tried using Figure 1 but the wavelet plane looked different.

Thanks

Ray

[Juan Conejero]

Hi Ray,

Hi Juan,
Would it be possible to get the same images of M33 that Vicent used in the Multiscale Wavelet Tutorial? When I try to download M33 from the Sky Survey my computer can't recognise the format.

http://www.pleiades-astrophoto.com/tutorials/LE/M33-vperis/en.html

Follow these steps:

1. Type M33 in the "Object name" field.

2. Click "GET COORDINATES". Wait while the page refreshes with M33's coordinates from the SIMBAD database. This should take a few seconds.

3. On the "Retrieve from" list, select "POSS2/UKSTU Red"

4. Type 60 in both Height and Width fields. Vicent's tutorial uses one square degree of DSS data.

5. Ensure that File Format is FITS.

6. In Compression, select "gzip". If you have a very-large-bandwidth connection, you may want to select "none".

7. Check the "Save file to disk" check box.

8. Finally, click the "RETRIEVE IMAGE" button. After some time (from a few seconds to several minutes, depending on many factors), you will be prompted where to save the file. By default, the proposed file name has no file extension - this is why your computer doesn't recognize the downloaded file. Ensure that the file name carries the proper file extension: .fit if you download without compression, .gz if you download with gzip compression.

My recommendation for step 8 is to discard the file name that the DSS sets automatically and replace it with something more readable like "red.fit", "green.fit" and "blue.fit" (or "red.gz" etc if you download with compression)

Repeat these steps for the three color channels. In step 3, select the POSS2/UKSTU Blue and Quick V images for the blue and green (visual) channels, respectively.

After downloading, you must register the three images with something like Registar, or an IP software that implements registration (PI still doesn't).

If you download gzip-compressed archives, you can extract the fits files with WinZip or WinRAR for example. Again, after extraction, the uncompressed files won't have file extensions. You must set them by renaming the uncompressed files, including the .fit extension manually.

Hope this helps. Let me know.
Juan

[Radar]

Cheers Juan,

I have been tied up all week with work. I will tackle this in the coming days.

Regards

Ray

[Radar]

Hi P Team,

I have been working with wavelets and getting good results.  

I have been generating large scale wavelet planes, then adding them back into the original image.

Sometimes I use curves and histogram on the wavelet plane, and other times I just add the wavelet plane straight back into the original image.

I'm curious as to the kind of processing that should be done to bring out detail such as in this example -

http://www.pleiades-astrophoto.com/examples/B142-vperis/en.html

Once the wavelet plane has been created are there standard or recommended ways to enhance and tweak the wavelet plane?

At the moment it is just trial and error for me, but I am very happy with the results I am getting, and it has been worth learning PixInsight.  

Thanks

Ray

[Juan Conejero]

Hi Ray,

I'm happy to know you're liking wavelets. They are among the most (or are the most?) powerful and versatile tools for image processing. Wait to see the new wavelet-based tools that we'll implement in the new Standard application. The present ones will seem toys.

I'm curious as to the kind of processing that should be done to bring out detail such as in this example -

http://www.pleiades-astrophoto.com/examples/B142-vperis/en.html

Once the wavelet plane has been created are there standard or recommended ways to enhance and tweak the wavelet plane?

I'll talk Vicent about this particular image to see if he can give you more information. I think that in the example above, he just used curves transforms with the large structures. Once you've isolated the large-scale structures, it is incredible what a good curve can do.

A very important thing is protecting the stars well, because they may suffer a lot during the whole process. A star mask can be built and loaded (perhaps combined with other mask) during critical processing steps. In this way the colors and smooth profiles that the stars always should have can be preserved.

Good news: we're implementing a specific tool for star mask generation. It will be available in build 203 of the Standard beta application, quite probably. With this build, we'll start publishing the new wavelet-based tools. The star mask generator is one of them.

Cheers
Juan

[Radar]

Hi Juan, sorry for my delayed response.

Cheers for the info.

Regards

Ray