M33 - Trying to Get my Head Around AP and Image Processing

[MercianDabbler]

I'm trying to do a bit better at AP and image processing. These are not my first attempts but previous images have used either just a single frame with a high ISO or stacking a very small number of frames and nothing more sophisticated than Picasa for image processing.

The frames were taken on 18th Nov in my garden. Clear Outside says I have Bortle 6 sky but I'm wondering if it is even that good.

M33 was pretty high compared to the other things I attempted on the same night. The photos were taken with an unmodded Canon EOS 70D through my ST80 piggybacked on my Vixen SP102 using the Super Polaris setting circles to find M33 and its RA drive for tracking. I could not see M33 visually through the SP102 (or my 10x 50s)... so I used the ST80 for the photos to give me a better chance of finding the target.

Reading up on ISO levels turned up a recommendation to use ISO 1600 with the 70D. I went up one 'stop' to 3200 because it was pretty gusty that evening.

31 light frames of 10s at ISO 3200 (more exposure was producing a very foggy looking frame which I thought did not bode well for pulling any detail out. Maybe that was an incorrect call but I needed to pick a setting and get on with it)

13 darks, 15 bias. The darks were done in between the lights because a neighbour's floodlight kept coming on so when the light was on I put the cover over the scope and took darks. It's likely that some of the lights were affected when the light came on in mid exposure.

The camera was left attached to the ST80 and the focus left alone when I took everything indoors.

Originally 20 flats done with the T shirt method the following morning but cut down to 2 because I suspect most of my flats had a gradient in them perhaps from being aimed at the sky too close to the house. The last few had a different aim but then it started raining causing me to duck inside for cover.

I did note the camera doing it's sensor cleaning routine when packing up in the evening... I wonder if this negates the dust fixing potential of taking flats the next day?

Dark flats... None. I've read about those but don't really understand the value of them.

Anyway, some pics. I'm not looking for others to do my processing for me thanks, just to get some pointers.

Stacking was done with the latest version of Deep Sky Stacker. I didn't mess with the settings.

The first attempt was done before I'd looked at any post procesing tutorials and with my pre-existing version of GIMP (2.8 IIRC) which will only deal with 16 bit TIFFs. The uncropped result after adjusting levels is below. The histogram looked very quantised so clearly 16 bit is not ideal. I had to reduce this image to 1/4 size before posting because the full size JPEG export wound up at 27Mb.

 

The most obvious issue with this is the amount of fogging in the lower right half of the frame, the peculiar colour cast of the whole scene plus the amount of coloured speckly noise in the background.

Then I installed GIMP 2.10 and worked with a 32 bit TIFF after reading some tutorials (mostly not about GIMP but a recent M31/M33 thread on here including GIMP and ImageJ). My steps were as follows...

• crop away the edges where the stack is not fully overlapped
• colour balance - use the levels tool and fiddle with mid point triangle. Initially tried to fully overlap the RGB histogram 'humps' but this was too much so then I did it less aggressively and eyeballed the resulting image.
• cloned the layer to start working on removing the skyglow
• despeckle at 15 radius
• use the heal tool to clone over the bright area left by the galaxy, cloning at right angles to the gradient
• Despeckle had left a lot of strange looking banding artefacts so I used gaussian blur with radius 60 to smooth these out
• use levels to temporarily pump up contrast to check the skyglow layer
• put skyglow layer above main image and set mode to difference, set opacity to 99% because it looked a bit too severe. I also tried subract mode but that looked more strange.
• merge skyglow and image layers
• lasso galaxy etc and apply unsharp mask, radius 1
• final crop
• export to JPEG at 98% quality

The background is a better colour so my first try at deleting the LP fog seems to have worked but I can't say that I'm hugely excited by the final result.

Any attempt to tweak it further with curves brings out some odd looking noise in the background. It also seems to have lost some of the outer parts of the galaxy that were visible in my much cruder first effort.

I did not use ImageJ. I read the other thread and it seemed to be used there (mostly??) for producing an artificial dark... but I had darks. Did it do something else that I ought to be doing?

Any pointers gratefully received!

[michael8554]

Processing gets tricky with that degree of gradient, maybe use a Light Pollution Filter on the camera ?

Take a lot more of those short exposures.

Turn off the DSLR Sensor Cleaning.

Yes you have lost the galaxy, try lassooing it in the first image and make it a new layer, apply Curves, then  you can merge into the de-gradiented backgound.

Michael

 

 

[AKB]

I'm certainly no expert, but I had a quick go at this.  There's probably more data there than you think.  I had to crop it a bit because there were still regions where stacking artefacts showed, which made it harder to deal with the background.  Really all you need is more of the same to make a good image.

Tony

 

Edited November 22, 2020 by AKB

[MercianDabbler]

43 minutes ago, michael8554 said:

Processing gets tricky with that degree of gradient, maybe use a Light Pollution Filter on the camera ?

Take a lot more of those short exposures.

Turn off the DSLR Sensor Cleaning.

Yes you have lost the galaxy, try lassooing it in the first image and make it a new layer, apply Curves, then  you can merge into the de-gradiented backgound.

Thanks Micheal.

I'd be interested to know whether folks think the gradient is purely from light pollution or if something else is going on. I've attached my first and last flat, again shrunk to quarter size.

Looking again, I'm not sure there was too much amiss with the earlier flats.

I guess it could be the floodlight - it wasn't lighting up the scope but it was lighting up everything from perhaps 10ft above ground level so I suspect there was some back scatter.

I shall definitely try again with more exposures and without any sensor shaking in between lgts and flats. TBH I thought that 30 frames was a lot. I was quite pleased just to find some evidence of M33 in the shot since I did not see it visually at any point in the process.

I might attach the camera to the SP102 next time. Now I at least have a photo of the star field so I should have a better clue how to find it even if I can't see it.

[MercianDabbler]

37 minutes ago, AKB said:

I'm certainly no expert, but I had a quick go at this.  There's probably more data there than you think.  I had to crop it a bit because there were still regions where stacking artefacts showed, which made it harder to deal with the background.  Really all you need is more of the same to make a good image.

Thanks Tony. That certainly looks better than my efforts. Any chance of knowing what steps you used?

[AKB]

Oh yes, of course,  Broadly speaking it was:

• crop the edges
• two consecutive goes at removing the background
• separate the stars from the nebula (Starnet++)
• noise reduction on the nebula
• stretching / colour balance / saturation (reduction) of both stars and nebula
• recombine the two
• final curves adjustment

This was, of course, on your originally posted image.  There should be further improvement available from the raw images and calibration frames.  Very hard to tell from your posted flats, but some of the background in your image is, no doubt, LP.

On your comment "more exposure was producing a very foggy looking frame which I thought did not bode well for pulling any detail out", this may not, in fact, be the case.  Some raw frames look extremely uninteresting until a bit of histogram adjustment is done.

Tony

Edited November 22, 2020 by AKB

[MercianDabbler]

35 minutes ago, AKB said:

On your comment "more exposure was producing a very foggy looking frame which I thought did not bode well for pulling any detail out", this may not, in fact, be the case.  Some raw frames look extremely uninteresting until a bit of histogram adjustment is done.

Thanks Tony. I was judging by the preview on the camera screen plus the histogram rather than the raw image. I'm used to raw images looking pretty dull. Looking at the histogram on its own it doesn't look too bad, the 'hill' being to the right of centre with no obvious sign of clipping. Maybe the longer exposure would have been better but having the whole frame in the preview looking so bright looked like the wrong answer to me.

When I've been out under darker skies the preview looks much more like the finished photo but I'm not really sure what is normal for doing lots of frames under LP.