Thar She Blows

[NigeB]

Hello All

I tried to take full advantage of the three whole nights of clear sky we had last month in these parts - so three nights leaving the observatory open all night hoping it didn't rain while I was asleep... I decided to concentrate on NGC 4631, the Whale Galaxy which is well placed right now.

Results below. All processing done in Pixinsight. I have to admit that while I'm quite pleased with the appearance of the image when seen from afar, I'm much less happy with the result up close. I've darkened the sky background more than I probably should, to try to conceal a multitude of sins.

Image information: TOA-150 at f/7, Mesu 200 mount, Atik 460ex at -15°C, Baader LRGB filters, Lodestar autoguider via a TS OAG.

L: 24 x 600 sec
R: 14 x 600 sec
G: 14 x 600 sec
B: 15 x 600 sec

Here are the main processing steps I applied:

1) Dark/Bias/Flatfield correction, alignment and stacking in Pixinsight using Winsorized sigma clipping.

2) Linear fit of R, G and B using L as the reference.

3) Applied a histogram stretch to all frames (transferring the Screen Transfer Function settings to make these permanent)

4) Performed Dynamic Background Extraction, subtracting the resulting background frames from images to remove some gradients and a few irritating flat field issues (some remain)

5) LRGB Combination performed

6) Increased saturation across the spectrum to bring out some colour. 

7) Found that star sizes in the red image are larger than the blue and green. This was giving an excessive red tint to all stars. So I went back to the red frame, applied a star mask, followed by Morphological Transformation to reduce the red star sizes a little.

8) Re-did the LRGB combination + saturation increase. Found the red halo issue was helped, but not completely removed.

9) Applied a star mask and reduced saturation levels in the red channel for stars only. This removed the general red tint and brought out some colour in the stars.

10) Some adjustment of the R G and B curves and cutoffs in Histogram Transformation to try and get a colour balance in the galaxy which looked right (aesthetically).

11) Final tweak of overall RGB histogram to hide the worst of the background and achieve what looked like sensible brightness, contrast and colour balance.

 

What I'm not happy with: the background is very noisy - lots of grain, lots of chromatic noise - zoom in and it looks terrible. Maybe a simple case of needing more data? The broader PSF in the red frame does not help - the star reduction process improved things a lot but I suspect is no substitute for a better match with L, G and B, and its effects are definitely still visible. It was caused either by a focus or a seeing issue (probably the latter). Lousy flat field - I relied on the first night's flats to correct the subsequent 2, and something had changed. Lesson learned. I could probably help to reduce the effects in post-processing but I think it's a second order issue compared to the other points above. Star shape is poor, possibly due to my heavily loaded mount (another thread on this elsewhere!)

 

Comments, criticism and suggestions very welcome. 

Thanks

Nigel

 

 

[Redscouse]

You know, as a newbie I often wondered why it was called the whale as I couldn't make out the shape. I see it now. Thank you. Great photo. 

[JAO]

Very nice image and excellent write up about the process of capture too. Love the colours , particularly in the HII regions of the galaxy.

[teoria_del_big_bang]

Nice image and a great in depth write up. Not that I yet understand all the write up as I still have lots to learn but it is great to read how these great images are made and even if a little bit of the info is digested it is a great help to anyone starting out in imaging.

Many thanks

Steve

[Allinthehead]

Hi Nigel. That's a very nice image. As you're looking for feedback the background can be fixed quite easily with Dbe and fixing your colour balance. There's a large green bias. Super details in the galaxy and i'd say your data is very good. 

One way to get a lot of tips and techniques from others would be to post your raw data. Lots of imagers are data poor this time of year. 

[wimvb]

Nice image and write up.

Personally, I wouldn't do a linear fit of colour data with L as a reference. Since L is a lot stronger than either colour, you are in essence pre-stretching the colour images to "catch up".

Try this:

• RGB combination of the stacked colour frames, without any other previous processing.
• DBE in the linear stage, check "normalize" in the correction panel of DBE
• Create two previews: 1 with just background, 2 with just the galaxy. Use preview 1 for background neutralisation, and 1 and 2 in colour calibration (1 as background, 2 as white reference), turn off structure detection.

These steps should result in a reasonably flat and neutral image to continue processing on. If you use Masked stretch as the first stretch for the RGB image, you also get more colour to work with in LRGB combination.

Btw, if you use large scale pixel rejection during stacking, you should have no trouble getting rid of that satellite trail. You can then have a less aggressive pixel rejection, which will increase SNR somewhat.

[NigeB]

Hi Everyone,

Many thanks for your positive comments / likes and constructive criticism! It is much appreciated.

Richard - that's very helpful: I must admit that I'd not really noticed the green tint (I think I was so relieved to get rid of the red, that I was happy with anything else!) but yes, now you point it out, it's really obvious. I'll have a go with DBE and adjusting balance. But I also like your final suggestion to make the data available - that's a good idea. So I've uploaded the entire file set to my Dropbox area. There's a lot of it... I've started a separate thread under the Image Processing Techniques section, here:

Wim - those are really useful suggestions which I'm going to try. Just so I understand correctly: you're suggesting an RGB combination at the beginning of the process; at what point should I add in the Luminance? Sorry: as you can see, I'm really at the bottom of this curve!

Thanks again

 

Nigel

 

 

 

[wimvb]

13 minutes ago, NigeB said:

you're suggesting an RGB combination at the beginning of the process

Yes, before anything else. You can do a linear fit of two of the colour images with the third as a reference, but I wouldn't use L for this. And it's not really necessary, as background neutralisation and colour calibration will take care of this much better than linear fit.

16 minutes ago, NigeB said:

at what point should I add in the Luminance?

You process luminance and colour seperately. L needs noise reduction, sharpening, etc to show all the detail you want. RGB needs lifting the colour, neutralising the background, etc. Noise reduction is easy: just blur the image slightly with convolution.

When both images are done, use lrgb combination. Just make sure that rgb has about the same background level, and similar contrast as L. If the two images are too much apart, lrgb will fail.

[NigeB]

On 01/05/2018 at 09:45, wimvb said:

You process luminance and colour seperately. L needs noise reduction, sharpening, etc to show all the detail you want. RGB needs lifting the colour, neutralising the background, etc. Noise reduction is easy: just blur the image slightly with convolution.

When both images are done, use lrgb combination. Just make sure that rgb has about the same background level, and similar contrast as L. If the two images are too much apart, lrgb will fail.

Wim, thanks for this and your posts on my other thread - very useful!

 

Nigel

 

[wimvb]

wrong thread.

Moved it here:

 

[Rodd]

On 4/30/2018 at 19:37, wimvb said:

Personally, I wouldn't do a linear fit of colour data with L as a reference. Since L is a lot stronger than either colour, you are in essence pre-stretching the colour images to "catch up".

Try this:

I agree--but, if you do want the lum linear fitted to the RGB--after you integrate the RGB, extract a lightness (in PI) or a luminance (in PS), and linear fit that to the lum, reinsert it into the RGB, then you can insert your lum which will be linear with the luminance channel of the RGB. (all this in non linear state btw).  If the histograms are identical--you can skip this step entirely.  That is one advantage of the STF in PI--

Rodd