M94-Please have a go

[alexbb]

Here's my quick go at your lum. I can have another go during the weekend.

Alex

[Rodd]

1 hour ago, Barry-Wilson said:

Rodd you certainly have the data as the very fine processing versions show.

It is a fiendishly difficult target to capture and process: as Rodd mentioned, my version is here, Barry's SmugMug site, and I plan to improve on this.

Not much room for improvement Barry--I suppose you could pile on the data (good data).  But I think yours is perfect the way it is.

Rodd

[Rodd]

39 minutes ago, moise212 said:

Here's my quick go at your lum. I can have another go during the weekend.

That looks superb.  I thought the core was overexposed as I could not fin any contrast--but you have shown that it is not.  Well done.  You know, I must ne missing something (a very great deal).  I have been over this data a number of times since the various versions starting appearing in the thread, and I still can't get it any where near what many of you have gotten.  

Thank you all for letting me know that my data is not the issue.  I was going to scrap it all and start again, but there is no need for that.  I will add data to what I have.

Rodd

[wimvb]

3 hours ago, Rodd said:

I thought the core was overexposed as I could not fin any contrast--but you have shown that it is not.

It most definitely is not. The core actually has some very nice detail. In PixInsight you can reveal this by using HDRMultiscaleTransform and a 50% mask. Or use LHE at various settings on clones of the image and combine these clones with Pixelmath.

[gorann]

On 27/04/2018 at 03:04, Xiga said:

Hi Rodd

Here's my take:

I found it quite tough to process i must admit, but not because of the quality of the data (which is excellent btw), just because RGB can be hard and i suspect this might be a tougher target than one might think. 

Here's a very brief run-down of what i did (the order might not be exact):

1. Combine the channels in Astro Pixel Processor. For this i assigned the R,G, and B to their respective channels, and i then assigned the Luminance and the R and G all to the Luminance channel. APP only lets you select a maximum of 6 channels at the moment, hence i wasn't able to use the G in the Luminance as well (perhaps a good thing, as it was the noisiest channel - although still far from noisy!). 

2. Gradient Reduction and star calibration also done in APP. Saved as a TIFF for finishing off in PS.

3. Multiple arcsinh stretching done in PS to preserve colour (this is still to be added to APP, otherwise i would have just used a DDP-stretched version from it). I don't have PI so i use @sharkmelley 's excellent method for doing similar in PS.

4. Highlight recovery via Shadows & Highlights to bring the core back a good bit. 

5. High Pass Filter with Soft Light to sharpen the core. 

6. Colour saturation using Jay GaBany's technique. 

7. Colour noise reduction using the Camera Raw filter. 

8. HLVG to remove the green.

9. Horizontal & Vertical banding reduction using Noel Carboni's actions.

10. Subtle curves to increase the contrast of the outer ring. Set the black point, and up the Vibrancy. 

11. Add a small amount of NR using Carboni's Deep Space NR. 

12. Crop out the slightly darker edges. 

I haven't done much RGB work so thanks for letting me play with it. 

Ciaran,

could you point me towards links describing sharmelleys "Multiple arcsin stretching" and Jay GaBaby's colour saturation? Would be interesting to try out.

[gorann]

Rodd,

I worked a bit more on your data, adding some more contrast to the core of the galaxy. Like Wim I have no data of my own coming in (until late August) due to the light nights up here at 60°N.

[ChrisEll]

Here's my attempt.

[Xiga]

3 hours ago, gorann said:

Ciaran,

could you point me towards links describing sharmelleys "Multiple arcsin stretching" and Jay GaBaby's colour saturation? Would be interesting to try out.

Here you go Goran:

Colour preserving arcsinh stretching:

https://www.cloudynights.com/topic/595610-photoshop-color-preserving-arcsinh-stretch/

 

Jay GaBany's colour saturation method:

http://www.astronomersdoitinthedark.com/dslr_llrgb_tutorial.php

You'll have to forward to the appropriate bit in Scott Rosen's video above. He also mistakenly refers to it as the Adam Block method, just so you know. 

Ps - I highly recommend recording both as actions so you can do them in a single click. 

 

 

[wimvb]

Still a bit soft, but here's my first pass at this data.

I wonder why we get such difference in colour from the same data. For this version I used pcc in PixInsight. I may go back and try traditional colour calibration.

[Rodd]

3 hours ago, gorann said:

Like Wim I have no data of my own coming in (until late August) due to the light nights up here at 60°N.

 

Looks good.  I have no problem posting data for you guys up in the twilight.  Its been pretty bad for me lately, but I image when I can.  Since switching to the ASI 1600 I have completed 2 images, M94 (not really complete without the lum), and M63--a whole LRGB data set (some problem with lum flats but the galaxy is unspoiled).  I also have oodles of KAF 8300 data.  You are welcomed to any of it should you need something to process.

Rodd

[Rodd]

1 hour ago, ChrisEll said:

Here's my attempt.

Allot of details in the central region.  The outer ring would be greatly improved with some proper lum.

Rodd

[Rodd]

14 minutes ago, wimvb said:

Still a bit soft, but here's my first pass at this data.

The outer ring is bright--but does not look overstretched....how?  The scale is markedly different--bigger.  Will cropping alone result in an increased scale, or do you need to upsample as well?

Rodd

[ChrisEll]

46 minutes ago, Rodd said:

Allot of details in the central region.  The outer ring would be greatly improved with some proper lum.

Rodd

Thanks, the outer ring kinda got inadvertently processed out a bit as I was concentrating on the galaxy!

[ChrisEll]

Rodd, one thing I'd say about the data is that it's very clean, I didn't need to do any NR on it and I only did a tiny smidgen at the very end for luck.

[gorann]

2 hours ago, Rodd said:

Looks good.  I have no problem posting data for you guys up in the twilight.  Its been pretty bad for me lately, but I image when I can.  Since switching to the ASI 1600 I have completed 2 images, M94 (not really complete without the lum), and M63--a whole LRGB data set (some problem with lum flats but the galaxy is unspoiled).  I also have oodles of KAF 8300 data.  You are welcomed to any of it should you need something to process.

Rodd

Very generous of you Rodd! I am sure we may take you up on that offer. Meanwhile is it ok if I post this one on my Astrobin with chared credits of course?

 

[Rodd]

2 hours ago, gorann said:

Meanwhile is it ok if I post this one on my Astrobin with chared credits of course?

Of course....meanwhile, I  am still pounding away at M94, trying to pick up some insight fro the wonderful posts on this thread.  Try as I might, to no avail.  I can get the central part looking good--with a decent background....but to get the outer ring like some of you guys did is eluding me.  I will keep at it (nothing else to do until I get some clear sky nights anyway).

Rodd

[gorann]

Thanks Rodd. You just need to get PS?. Or what do you say Wim? In PS you just need to use curves to boost the brightness level of the outer shell.

[Rodd]

5 hours ago, ChrisEll said:

Rodd, one thing I'd say about the data is that it's very clean, I didn't need to do any NR on it and I only did a tiny smidgen at the very end for luck.

Then I am definitely missing something.  I find a graininess that I call noise--I just don't understand.  Bur Wims image looks amazing even at fulll resolution.  I don't know what step to take next

[gorann]

22 minutes ago, Rodd said:

Then I am definitely missing something.  I find a graininess that I call noise--I just don't understand.  Bur Wims image looks amazing even at fulll resolution.  I don't know what step to take next

Since you do it in PI, someone like Wim could figure it out and hopfully it is easily solved

[wimvb]

53 minutes ago, gorann said:

Since you do it in PI, someone like Wim could figure it out and hopfully it is easily solved

apparently, I did. 

2 hours ago, Rodd said:

Of course....meanwhile, I  am still pounding away at M94, trying to pick up some insight fro the wonderful posts on this thread.  Try as I might, to no avail.  I can get the central part looking good--with a decent background....but to get the outer ring like some of you guys did is eluding me.  I will keep at it (nothing else to do until I get some clear sky nights anyway).

Rodd

A lot of noise reduction went into this image, actually.

In the linear stage tgv denoise on L, and MMT (8 layers) on chroma, using a combination of Jon Rista's method and this:

https://wimvberlo.blogspot.se/2016/07/noise-reduction-for-dslr-astroimages.html

Masked stretch as the main stretch for the rgb data. I decreased the clipping factor from default 0.005 to 0.0025

Deconvolution of the L data, noise reduction (TGV and MMT on L)

To get the ring right, use a range mask that targets just the ring. Then apply Exponential Stretch. Finish off with another round of noise reduction.

[Rodd]

6 minutes ago, wimvb said:

apparently, I did. 

A lot of noise reduction went into this image, actually.

In the linear stage tgv denoise on L, and MMT (8 layers) on chroma, using a combination of Jon Rista's method and this:

https://wimvberlo.blogspot.se/2016/07/noise-reduction-for-dslr-astroimages.html

Masked stretch as the main stretch for the rgb data. I decreased the clipping factor from default 0.005 to 0.0025

Deconvolution of the L data, noise reduction (TGV and MMT on L)

To get the ring right, use a range mask that targets just the ring. Then apply Exponential Stretch. Finish off with another round of noise reduction.

There are allot of ways to do it apparently--someone above didn't use any noise control and his image looks pretty good. Not as good as yours perhaps--but much better than mine. I will have to look into your noise reduction method.  However, I have  never have good results with masked stretch.  I have tried many times. I tried ET on the ring.  Looked terrible.  When I used more noise control I ended up with a smeared mess that still had a graininess to it.  If I have to use more than a tiny amount of noise control something is wrong.  

I have never had much luck with deconvolution in linear state--or with TGVD in linear state.  But, right or wrong, both of these work very well for me in non linear state.  But those are the finishing touches.  With this image, I can't get past CC.  I found the Photometric Color Calibration works better on this image than the old BN and CC.  But that is as far as I can process this image--beyond that, it degrades no matter what I do.  

[wimvb]

Noise reduction can be tricky. Here's a version with bn+cc but too much noise reduction. Always use noise reduction with a mask to protect areas with high SNR.

The trick with good deconvolution is getting deringing and regularisation right, and not expecting too much from it. Try to keep the improvement subtle. Then use HDR methods in the nonlinear stage to get an increase in local contrast.

[wimvb]

3 hours ago, gorann said:

Thanks Rodd. You just need to get PS?. Or what do you say Wim? In PS you just need to use curves to boost the brightness level of the outer shell.

Why keep it simple, when you can complicate things? 

[Xiga]

Thanks again Rodd for posting this. The season is pretty much over for me now until late August, so i've enjoyed working with this.

Also, i decided to go back and add another subtle curve, to target just the outer ring. As expected, it did of course increase the noise around the edge of the outer ring, so i applied some NR just to these specific areas (using a layer mask). On this occasion i used the NR from the Camera Raw filter, just for a change - it's surprisingly good i find. 

As you can see, there are many ways to skin this cat!

 

[Rodd]

1 hour ago, wimvb said:

Noise reduction can be tricky. Here's a version with bn+cc but too much noise reduction. Always use noise reduction with a mask to protect areas with high SNR.

The trick with good deconvolution is getting deringing and regularisation right, and not expecting too much from it. Try to keep the improvement subtle. Then use HDR methods in the nonlinear stage to get an increase in local contrast.

The colors I get using BN and CC are nothing like this.  I get a muddy pinkish gray.  I always use an extracted lightness channel for local support when I use TGV denoise--I saw that in a tutorial.  I will try the decon in linera agin followed by HDR in non linear--I don't think I tried that particular combination before.