Another NGC 253 - Ho Hum...

[pvelez (Pete)]

My latest target is NGC 253 - again. I think I have imaged this every year for the last 3 years. There have been improvements in image quality and my processing skills each time but...I am never really happy with the result.

This year's effort is attached. It only contains 4 hours of Lum and 4 each of R, G and B. All 5 minute shots unbinned. Taken with PW CDK 12.5 and FLI PL16803.

Also attached is a version with Ha added. I am sure it is too scarlet for some (many!) - what pleases me about this is that in trying to keep the red under control, the image is less saturated and stretched less.

Here are links to larger resolution images:

http://www.pbase.com/equitius/image/161593014

http://www.pbase.com/equitius/image/161593017

Both can do with another 3 or 4 more hours of Lum - hopefully this week will remain at Coona so I can collect this.

I am sure these would look better with more data and so I should combine work over the last few years. Its worth giving this a go I reckon.

As always comments and criticisms are welcome - all processed in PI so if you have thoughts on how to refine the images with PI, definitely give me your views.

Pete

[codemonkey (Lee)]

Nice work Pete! I like the +Ha version especially. The colour is very pleasing.

Looks like a strong gradient in the background and maybe too much noise reduction for my tastes. I'd probably use dynamic background extraction to fix up the gradient, dial back the noise reduction and maybe pull the background a bit (may not need so much noise reduction then).

[strongmanmike (Michael)]

Each is a fine result (on this fast becoming waaaay over imaged galaxy ) Pete and any opinions on which looks better are really only going to be purely personal. My only constructive critiism is the noise reduction looks a tad unatural...buuut I guess that is a persoanl opinion too ...

Great stuff.

Mike

[alpal]

Hi Pete,
A nice job with 20 hours of data!
I really like the Ha version - it makes the image pop.
I agree with others that it looks overly smooth &
also I would have selectively added more contrast to the core of galaxy.
I only wish my own images were as good.

cheers
Allan

[Placidus (Mike and Trish)]

Hi, Pete,

Very nice indeed, with excellent underlying image capture. Nice level of colour saturation.

I'd suggest clipping the zero point quite a bit closer to the foothill of the histogram, reducing the blue a tad and increasing the red a fraction, and doing a teensy bit of wavelet sharpening.

On such a bright target, you've gotten away with the 5 minute subs, but in general I'd suggest going to something like half an hour with the 16803 on a 12.5" CDK.

Lovely work.
Best,
Mike

[DJT (David)]

Nice Pete. Big fan of the one with the excess of red.

[E_ri_k (Erik)]

Looks great Pete I prefer the HA version, great job

Erik

[RickS (Rick)]

Hi Pete,

You should be able to get some great detail with that setup. I presume the Pbase images are downsampled? They look nice but not as sharp as I'd expect.

If you have previously captured data with the same image scale then that should be fairly easy to add into the mix. Just calibrate with matching bias, darks and flats then register (or re-register) everything against the same reference image and integrate together. You can even mix image scales but adding low res data may not improve the overall image.

Cheers,
Rick.

[Paul Haese]

I like the one without the Ha the most Pete. My recent reprocess of 253 has a very similar colour to yours which I think is indicative of it being perhaps close to the correct colouring for me.

[pvelez (Pete)]

Originally Posted by codemonkey

Nice work Pete! I like the +Ha version especially. The colour is very pleasing.

Looks like a strong gradient in the background and maybe too much noise reduction for my tastes. I'd probably use dynamic background extraction to fix up the gradient, dial back the noise reduction and maybe pull the background a bit (may not need so much noise reduction then).

Thanks Lee

I'll have another look at the gradient. I usually apply DBE to the masters just before combining but I may have skipped it on this one.

NR has been a steep learning curve for me - perhaps I've been overzealous. When working full screen at native resolution every little gnarly region looks ugly, hence the smoothing. I have applied MLT to both the Lum and RGB - would be interesting to see if I skipped the NR on the RGB component.

Originally Posted by strongmanmike

Each is a fine result (on this fast becoming waaaay over imaged galaxy ) Pete and any opinions on which looks better are really only going to be purely personal. My only constructive critiism is the noise reduction looks a tad unatural...buuut I guess that is a persoanl opinion too ...

Great stuff.

Mike

Thanks Mike - I agree, NGC 253 is sooooo last season. (Though I do have something a bit less well exposed to post when I collect enough data).

Yep - will dial back the NR

Originally Posted by alpal

Hi Pete,
A nice job with 20 hours of data!
I really like the Ha version - it makes the image pop.
I agree with others that it looks overly smooth &
also I would have selectively added more contrast to the core of galaxy.
I only wish my own images were as good.

cheers
Allan

Thanks Allan - I souped up the Ha this time. The script for adding narrowband in PI allows the user to set the percentage of blending. The default is 120% which I usually reduce to about 50%. This one is the full 120% - like turning the amp up to 12 I guess.

I applied a local histogram adjustment to the core but I think you are right 0 it could do with a bit more.

Originally Posted by Placidus

Hi, Pete,

Very nice indeed, with excellent underlying image capture. Nice level of colour saturation.

I'd suggest clipping the zero point quite a bit closer to the foothill of the histogram, reducing the blue a tad and increasing the red a fraction, and doing a teensy bit of wavelet sharpening.

On such a bright target, you've gotten away with the 5 minute subs, but in general I'd suggest going to something like half an hour with the 16803 on a 12.5" CDK.

Lovely work.
Best,
Mike

Cheers Mike

Sub length is an interesting one. Running a few different optimum sub exposure calculators with my set up, about 4 minutes is the general recommendation. I have gone out to 10 minutes but not much beyond that except in Ha. I might have a play with that.

Originally Posted by DJT

Nice Pete. Big fan of the one with the excess of red.

Thanks David

Originally Posted by E_ri_k

Looks great Pete I prefer the HA version, great job

Erik

Thanks Erik

Originally Posted by RickS

Hi Pete,

You should be able to get some great detail with that setup. I presume the Pbase images are downsampled? They look nice but not as sharp as I'd expect.

If you have previously captured data with the same image scale then that should be fairly easy to add into the mix. Just calibrate with matching bias, darks and flats then register (or re-register) everything against the same reference image and integrate together. You can even mix image scales but adding low res data may not improve the overall image.

Cheers,
Rick.

Spot on Rick - the PBase images are resampled at 50%.

The images do look soft - I have wrestled for a while now trying to both smooth the background and sharpen the target. There seems to be a trade off here and I don't have it quite right yet.

My NR of choice at the moment is MLT using a linear mask. This produces some blotches that are smoothed by ACDNR when I take the image non-linear. So its a double NR step that takes out some of the detail. Then I apply MMT to sharpen - masked with a Lum mask. My challenge is to sharpen the target while not overly accentuating the stars which become far too prominent - I've tried to reduce the stars MRT and then sharpen but it doesn't always work.

I expect the sensible thing to do is to subtract a decent star mask from the Lum mask and then use that to protect stars and background when sharpening - is that your approach? Of course that requires a precise star mask - my learning edge I'm afraid.

Of course Decon could tighten the image up when linear - but I usually end up with worms and Mr Sidonio, Resident Decon Monitor would spy them in a flash.

Originally Posted by Paul Haese

I like the one without the Ha the most Pete. My recent reprocess of 253 has a very similar colour to yours which I think is indicative of it being perhaps close to the correct colouring for me.

Cheers Paul - its good to hear that the colour matches your work. I do love increasing the saturation and sometimes it goes a bit too far.

Pete

[RickS (Rick)]

Originally Posted by pvelez

My NR of choice at the moment is MLT using a linear mask. This produces some blotches that are smoothed by ACDNR when I take the image non-linear. So its a double NR step that takes out some of the detail. Then I apply MMT to sharpen - masked with a Lum mask. My challenge is to sharpen the target while not overly accentuating the stars which become far too prominent - I've tried to reduce the stars MRT and then sharpen but it doesn't always work.

I expect the sensible thing to do is to subtract a decent star mask from the Lum mask and then use that to protect stars and background when sharpening - is that your approach? Of course that requires a precise star mask - my learning edge I'm afraid.

I use MLT with a linear mask before stretching as my initial noise reduction step. I'm not a fan of TGVDenoise. I still use ACDNR for chrominance noise reduction but not on my luminance. If additional nr is required after stretching I use MLT but with a conventional clipped luminance mask.

Yes, a clipped lum mask with stars removed is useful for sharpening. You don't need to be too careful with the star mask if you use a little PixelMath trick: try an expression like "iif(star_mask>0.1,0,$T)" and adjust the 0.1 up and down as needed. Soften the mask a little with Convolution or by removing a couple of wavelet layers afterwards.

Cheers,
Rick.

[Peter Ward]

Nicely done. The hues are spot on, though a little too pastel (ie purely a brightness function) looking on my monitor.

plus.... you can ever have too many NCG253's

[RB (Andrew)]

Beautifully done Pete.
I like both, maybe prefer the one without the Ha, but it's a close call for me.
Very nice work !

RB

[pvelez (Pete)]

Originally Posted by Peter Ward

Nicely done. The hues are spot on, though a little too pastel (ie purely a brightness function) looking on my monitor.

plus.... you can ever have too many NCG253's

Thanks Peter

Originally Posted by RB

Beautifully done Pete.
I like both, maybe prefer the one without the Ha, but it's a close call for me.
Very nice work !

RB

Cheers RB

[pvelez (Pete)]

Originally Posted by RickS

Yes, a clipped lum mask with stars removed is useful for sharpening. You don't need to be too careful with the star mask if you use a little PixelMath trick: try an expression like "iif(star_mask>0.1,0,$T)" and adjust the 0.1 up and down as needed. Soften the mask a little with Convolution or by removing a couple of wavelet layers afterwards.

Cheers,
Rick.

So I've been playing with this expression Rick - what does it actually do? I find it simply creates rings around the star. That can't be hat you intend is it?

Pete

[RickS (Rick)]

Originally Posted by pvelez

So I've been playing with this expression Rick - what does it actually do? I find it simply creates rings around the star. That can't be hat you intend is it?

Pete

You need to apply PixelMath to the clipped Luminance mask (and change the name of "star_mask" if the star mask happens to be called something different.)

It should set values in the Luminance mask that correspond to stars in the star mask to zero, i.e. make black holes where the stars were.

[pvelez (Pete)]

The PI team should call it Pixel Magic not Pixel Math! Thanks for that. Something new to fiddle with - there is always something to learn in PI

Pete

[RickS (Rick)]

It is indeed a very powerful tool, Pete! I do lots of stuff with it including manipulating masks (need a blue star mask? no problem...) blending images (all the Photoshop blending modes are possible) and even perfect removal of satellite trails.

[Shiraz (Ray)]

nothing hohum about these Pete - very nice images.