ESO critique of my NGC 253 image.

[sjastro]

From the "NGC 253 imaged as a dim Dwarf galaxy in Colour" thread.



ESO's response as follows. My image marked out by ESO and referenced in the response is in the attachment.



Olivier is looking at this image in the cold hard light of the professional astronomer, it is encouraging he is prepared to invest further time in this image and hasn't rejected the extended halo out of hand.

Steven

[Placidus]

Excellent discussion, Steven, and once again a stunning image.

Not thinking about your image specifically, but just thinking about general mechanisms, what might create a very wide artificial halo around any object? I can think of four. There might be others.

The first is actual physical smearing of light from the main galaxy by the optical path. Dust in the air, fingerprints on the optics, reflections off shiny bits. Perhaps only a thousandth part of the light from the galaxy needs to get spread to create a detectable false halo. You could perhaps check this experimentally by photographing Rigel, and seeing what the background looks like when very strongly stretched.

The second, as your ESO colleague mentioned, is any kind of low pass filtering operation (smoothing). For example, a wavelet filter with too many layers. To spread say 400 pixels, you'd have to use a wavelet filter with 8 or 9 layers. You could check by applying the same filtering to the Rigel shot, or by reprocessing with fewer wavelet layers.

The third general mechanism (unlikely to apply in the current case) is superb but not quite perfect flats. Residual directionality in the test light used to produce the flat (or not having exactly the same optical path) can produce quite complex gradients. We're only looking at an error of a fraction of one percent Rephotographing with the object off to one side would resolve that issue.

The final one I can think of is the use of a gradient removal tool. This can effortlessly produce a DARK halo around a galaxy. Perhaps a complicated gradient removal tool can produce a dark halo further out, causing an apparent light halo inside it. The only way I can think of to test this is to do the processing without a gradient removal tool.

Once again, a superb photo, and I don't believe I'm directly addressing your image, but this kind of 'systems engineering' approach might be useful to anyone exploring undocumented super-faint, extensive haloes.

Very best,
Mike

[sjastro]

Thanks for your comments Mike.

The large halos in my image are mainly caused by the application of a star mask that was too large for the star. Stretching the image results in everything but the region underneath the mask being stretched.

Of greater concern was the alternating brightness variation as one moves away from the disk of the galaxy.
This was caused by an incompatibility between the skyglow subtraction routine and the non linear stretching.
I believe I have fixed this problem by modifying the routine.

At this stage I have sent Olivier a revised version without the skyglow subtraction routine. The halos and brightness variations are gone but the galaxy halo is not as conspicuous as in the original image.
I suspect ESO will measure the pixel values from the edge of the disk to the edges of the frame. A drop off in the pixel values down to the background noise is probably a strong indication the outer halo is real.

If this is the case the next step would be to submit an image where the skyglow has been subtracted.

The interesting point in the discussions with Olivier is the "cultural differences" between amateur astroimagers and their professional counterparts.
We amateurs convey information through the visual presentation of our images, the professional on the other hand extracts quantitative information from the image.
The more we amateurs process an image, the less useful information is available. This was certainly the case when I performed a noise reduction on the image!

Hopefully this exercise will conclude where an image of an obvious extended halo is found to be real and is backed by scientific information.

Regards

Steven

[LightningNZ]

This is such cool work Steven. Thanks so much for sharing the process with us, it's fascinating. :)
-Cam

[clive milne]

I respect what you are doing here Steve.
I would be interested to know from Olivier if colour data using ubvri bands would be more useful (valuable)
I imagine it would be relatively straight forward to script regular calibration frames utilising standard reference fields during an imaging run with the data tagged to each exposure. They might even have the script to give us ex-gratis.
I doubt the throughput would be as high as you would get with high transmission lrgb filter sets, but it is a trade off that would be worth making in certain circumstances imho.

Anyway, there is no harm in putting the idea to the ESO guys.

At the very least, they will no doubt take you/us more seriously if we develop a more rigorous protocol with respect to our data acquisition.

Nice work btw.

Best
c

[sjastro]

Thanks Clive.
I'll pass on your question when I get a response to the latest image submission.



Thanks Cam.

[Shiraz]

:thumbsup: nice to have some pro feedback - very interesting work as well.

[sjastro]

Thanks Ray.
The feedback throws a whole new light on processing.

Steven

[sjastro]

Hello Clive.
Here is the response from Olivier.

[clive milne]

Thanks for following that up Steve. ..

Incidentally, it occurs to me that the variation in spectral sensitivity across the gamut of ccd sensors (available) renders strict adherence to Johnson filter curves a redundant and possibly even counter productive exercise if each exposure is going to be calibrated against reference stars anyway.

best
c

[sjastro]

The good news is ESO is using science in evaluating whether the extended halo is real or not, instead of a subjective yes or no response based on looking at a pretty picture.
Their emphasis is on the noise in the background rather than the image itself.

The bad news is my processing efforts are hindering their efforts in making that evaluation.
I am becoming aware of the considerable differences in how amateurs and professionals process images.

Olivier Hainaut from ESO wrote...


This was the second image submission to ESO where I took pains to minimize the effects of noise smoothing.
Evidently a mild degree of wavelet processing in PI with the noise reduction option "on" was enough to wreck ESO's attempts in evaluating the noise in the regions as shown in the attachment.
What I suspect ESO were investigating is the amount of Poisson noise in the region of interest. This is "good noise" related to the statistical variation of real data. If "bad noise" or Gaussian noise related to factors such a read out noise and noise produced by dark and flat field subtraction dominates, then the halo is nothing more than an artefact that has been enhanced by processing.

The next submission if ESO are still prepared to tolerate me:), is an image where there has been no processing except stretching including the pixel mapping process.
Hopefully ESO will also provide more information on how they evaluate the background noise.

Incidentally the latest high resolution image is here.
http://members.iinet.net.au/~sjastro/ngc253_colour_halo.jpg

Regards

Steven

[sjastro]

I have been able to isolate and measure the Gaussian noise contributions in the areas highlighted by ESO for investigation.

The Gaussian noise was measured in the darks and flats used in the image reduction process for the luminance image. The darks also include bias noise.
The Gaussian noise from stacking and stretching the image was also evaluated. No sharpening was performed.

Using this information the percentage of Poisson noise in the luminance image was calculated.
Poisson noise from extraneous sources such as light pollution, vignetting, hot spots and field flooding were eliminated as possibilities.
Therefore the Poisson noise was from the external halo itself.

In region 1 (the area above ngc 253 in the attachment), the percentage of Poisson noise in the total noise was 79.6%.
In region 2 (below ngc 253) the percentage was 81.8%.

As these percentages are high I believe the external halo is real.
Hopefully ESO will confirm this by their own noise evaluations on the same image.

Regards

Steven

[Bassnut]

Having look at your excellent rendition, I stretched the hell out of my effort (Lum). The halo is certainly real, its huge!. I didnt apply any processing apart from uber stretch. Noisey as hell, not as smooth as yours.

[DavidU]

Very interesting stuff !

[sjastro]

That's a good effort Fred.
What was the total exposure time for your luminance?

I had no doubt the halo is real but convincing an astronomer without quantifiable data is futile. ;)


It certainly has been David.

Regards

Steven

[Rex]

Steven, this is great work your doing and I find this thread very interesting. Watching and waiting as I'm sure are many others.

[alpal]

Hi Steven,
I doubt that such a large halo is real - I rely only on skeptical scrutiny.

I tend to trust the 32" Chilean Advanced Robotic Telescope pictures as a guide.
see here:
http://www.chart32.de/images/objects/galaxies/NGC253/NGC253-80P.jpg

Yes - the halo does extend much further than most amateur pictures
but is no where near the size of your picture.

cheers
Allan

[Bassnut]

6 hrs, 20min subs. I may have strectched it too far. Given the halos around big stars, which arnt real I suspect, the main halo might also mostly be fake, although I dont know how that happens.

[sjastro]

Thanks Rex.
Olivier has been somewhat quiet lately.
I know one of his projects is the Earth based observations of comet 67P/Churyumov–Gerasimenko which Philae landed on recently, so perhaps he has been occupied with much more important events.;)



Alan there is absolutely no doubt the extended halo is real.
It's been shown in Mike's negative image (with the mysterious S-bend), Octane's and now Fred's.
It shows up in professional images in radar, X-ray and FUV/NUV as shown in the attachment.
Since the halo emits synchrotron radiation in NUV and is a continuous rather than an emission spectrum, the radiation spills over into the visible range.
ESO's role in this is to play the devil's advocate, they know the halo is there, but they can't make a judgement call until they examine an image where the background noise has not been altered in any way by processing.



My luminance exposure was only 1.5hrs exposure total.
I think that has added to the complications.
If it was a 30 hr exposure like the Carina dwarf galaxy, I don't think ESO would have any doubts.:)

Regards

Steven

[alpal]

Steven,
Hi Steven,
you seem to be very certain.

Can you explain why - in the Chilean picture here:
http://www.chart32.de/images/objects/galaxies/NGC253/NGC253-80P.jpg

there are very faint galaxies showing up that are not in your image yet
you have a halo showing instead?
Surely if the Chilean telescope picked up such faint galaxies it would pick up the halo too?

cheers
Allan

[sjastro]

Are you sure about that Allan?
Perhaps we can narrow down the comparison by looking at the region around the stars HD4555 and HD4572 (the two bright stars above the disk).

Did you use my higher resolution image for comparison?
http://members.iinet.net.au/~sjastro/ngc253_colour_halo.html

You also have to also consider that three other astroimagers in this forum have produced more or less the same structure and the halo has been observed at other wavelengths.

The best method for comparison is to reprocess the raw data set from the Chilean scope so that the only differences are in the processing methods.

This was done in the Carina Dwarf Galaxy thread.
ESO provided me with their raw data which I reprocessed and sent back to them.
The comments were most encouraging claiming the reprocess resulted in an image with greater depth.
The only negative was the non linear stretching prevented them from using the image for photometric purposes.

Regards

Steven

[strongmanmike]

It is very well known that NGC 253 has an extended halo, I am unsure of what all the fuss is about? Is it that you feel it is bigger and more extensive in your data Steve?

Mike

[sjastro]

Mike,

I think even it's even larger than the FOV of my ST-X10ME.
I'll revisit this one with my STXL-11002 and a much deeper exposure.

Regards

Steven

[alpal]

Steven,
Hi Steven,
Ok but maybe at the top right hand corner at 1 o'clock where you still have a halo would be better.
I have made a comparison image between yours & the Chilean one.
I hope they don't mind - it's just a tiny section.
I haven't equalised the background brightness - it's as is.
Can you see all the little smudged galaxies coming through on the right hand Chilean photo
but not on your left hand one?

cheers
Allan

[sjastro]

Alan,

The 32" is used in seeing conditions that can be as good as 0.3 arcseconds.
Compared to my site, this is about 10X better then what I can achieve.
Also the 32" has an aperture 3X larger than my own scope.

The small galaxies visible in the 32" are probably beyond the resolution limits of my scope and location site and can therefore be treated as point sources.
As point sources their visibility in an image is dependant on the aperture and exposure time.

The halo on the other hand is not a point source but an extended object and depends on the f/ratio and exposure time.
So you are not comparing apples with apples.

The other issue is how each image is processed.
Clearly the 32" hasn't been processed with the idea to bring out faint detail as illustrated by the inky black background and high contrast.

Regards

Steven

[alpal]

What if I equalise the background of the Chilean pic with yours?

[rustigsmed]

I Al, I think the point Steven has been raising throughout the thread is the imaging processing techniques used. ie making a pretty pic vs linear stretching. unless you have the raw files from the Chilean image you can't compare data sets.

cheers

rusty

[alpal]

Yes - you have a point.
I try however as attached.
I removed all the colour & gave a 5 pixel blur to the Chilean image.
The faint galaxies still stand out better:

[sjastro]

I wrote.


Olivier Hainaut ESO replied.


The noise in my image has become more important than the image itself, particularly the Poisson noise from the sky background.

As yet my last submission has not been evaluated.


This has been a real eye opener in how professionals deal with astro images.

Regards

Steven

[sjastro]

Since flat fields are a source of Gaussian noise (in particular when poor flats are used), the subject of how professional astronomers perform flat fields cropped up.

Olivier Hainaut wrote.

[RobF]

Wow! That's incredibly useful info. Worthy of a thread of its own :confuse3:
Makes perfect sense when you think about it. Probably also emphasises the increasing difficult in obtaining good flat fields with increasing focal length most people encounter.

[sjastro]

I read a paper where Dome flats are considered to have an all round superiority to twilight and sky flats when LEDs are used as the illuminating source and a screen material that efficiently reflects light over a wide range of wavelengths.



Ref. http://arxiv.org/pdf/astro-ph/0510233v1.pdf

Regards

Steven

[RobF]

Thanks Steven. Will have a read.
I subscribe to any information that reinforces the notion there is a scientific reason why the flats were so bad for my old 8" Newt (any reason except "myself" that is :))