Help me improve this M42 image!

[Matthew.Blake]

OK, trying to use my RAW files properly I've finally got Registax to load them (well, TIFFs of them). I wanted to use DSS, but it keeps saying "Out of memory" now. Sucks.

I've got a stacked image that is at last better than any one of its component parts. Hoorah. Except in one regard, there is a bit of a problem with LP that I can't remove without also killing detail in the nebula. The first image retains all the finer detail but has the excess LP, the second gets rid of the LP but loses the fine detail. The image was made with 10x30second shots, USB extension cables arrived today to help me take longer shots with EOS utility but it's snowing tonight :-(. I have added no flats, darks or bias frames to this (not fully understanding them yet). Is this an example of the need for a flat frame so Registax can deal with non-homogenous LP?

[Tim]

Hi Matthew.

What settings are you using to make DSS conk out? It might be an idea to just stack a section of the image by dragging the red rectangle around it. RAW files from dslrs can be rather big. Make sure you dont have the drizzle options checked.

To get rid of the light pollution from this image you could try the free version of Pixinsight (LE), just google pixinsight and you will come to it. Read the section about the background removal tool.

Try if you can to learn about flat frames. They will help you to get rid of the vignetting (dark areas around the edge) from future images.

What telescope are you using? Ideally the brightest area of the picture will be fairly central, rather than offset as above. Sometimes this is down to collimation, and sometimes due to the camera or adapters not being squarely fitted. If you cant fix that, then on your next target try to position the main detail near the centre of the offset, and crop away the unwanted stuff.

Its a good start btw.

[Euan]

OK, trying to use my RAW files properly I've finally got Registax to load them (well, TIFFs of them). I wanted to use DSS, but it keeps saying "Out of memory" now. Sucks.

Have you got Drizzle enabled? This really chews up memory

[Matthew.Blake]

I took the picture through a Skywatcher 200p on an EQ6.

What do you know, it was drizzle chewing up my memory. I've only got 4GB! Good grief.

[roundycat]

Matthew,

you have a good start here but the image is severely clipped at the dark end. People talk about the black point all the time but there should not really be any black. I tweaked the mid point to get the re-posted version.

Flats correct optical system defects, they will have no effect on light pollution.

Dennis

[Deneb]

Hi Matthew

Here's my 2p's worth with your image, but more subs, Bias & Flats needed next time

Cheers

Nadeem.

[Matthew.Blake]

Hi Matthew

Here's my 2p's worth with your image, but more subs, Bias & Flats needed next time

Cheers

Nadeem.

Nicely done, you're going to have to tell me what you did to it though so I can learn how to do it myself?

I'm moving on to flats, darks and bias frames soon, took me a while to figure out how to stack something that wasn't an 8bit/ch JPEG.

[Deneb]

Well, to get rid of some of the Gradient I used gradient exterminator, I re adjusted the levels so the black was not clipped that much, adjusted the histrogram. A little adjustment in curves, did not bother using brightness & contrast, used unsharp mark, to give the stars a bit more definition, then ran noise reduction. Also a pinch of salt

Build yourself a lightbox for your flats.

Cheers

Nadeem

[LukeSkywatcher]

They all look GOOD to me but i think Deneb's version has nailed it.

[simon hicks]

Hi Matthew,

I had a quick go with your image. The big problems I saw....which agrees with the other posts....is that there was really big gradients in the background and the background had been clipped.

To get rid of the background you could use the Gradient Exterminator (is that in PhotoShop?) or the DynamicBackground removal tool in PixInsight, or a number of other variants. Try to do this early in the processing, definitely before any clipping. Clipping deletes data from the image and you really want to save all the data till maybe almost the final stage if you can. I must admit I do usually clip some of the data....but never eating into the big median curve.

There is some annoying green in the dimmer parts of the image....but I'm sure this is due to clipping the data before removing the gradients. If you reprocess from the stacked image then I'm sure that will disappear.

And looking at what is left in the image, then it indicates that you've actually got some great starting data....and there is a lot more in there to be extracted....as long as you haven't thrown the RAW files away!

Cheers

Simon

[Matthew.Blake]

Ok, here's my attempt at adjusting the curves. I've included the curves too. How is this for an improvement? I've no access to gradient exterminator sadly.

[ollypenrice]

The team have all done a good job on this one! The first image seems to me to have more of an issue with vignetting through lack of flats than with LP but I could be mistaken.

Olly

[michael.h.f.wilkinson]

Hi,

I have just tried some new image processing software I wrote to kill the background gradients, without distorting the other details. The method is called a k-flat attribute filter, in which you specify the properties of objects you wish to remove (in this case anything with an area larger than 102400 pixels). It also has some noise suppression performance. I thinks this looks good. Comments welcome. If people are interested, I can write a GUI-based windows version (this is just a command-line program).

The method has just been accepted for publication: http://www.cs.rug.nl/~michael/OuzounisWilkinsonPAMI2010fpage.html

Note that I did not adjust ANY curves, to show what the filter does.

[Matthew.Blake]

I shall have a read of your paper, thank you.

[simon hicks]

Hi Matthew,

I fear you will not get any more out of the wonderful data you have UNLESS you get rid of the background gradients first.

Also, your Curves shows how you have clipped the red, green and blue to try and balance them. This the wrong way to do it and has resulted in the green haze seen in the version that I posted. By clipping it like this you have changed the colour balance bwtween the bright parts and the dim parts of the image. Hence the bright parts of the nebula look great, but the background is ripples of colour.

If curves is all you have to balance the colour then try to keep the lower clip point at zero and the upper clip point at 1 and just drag the central part of each curve diagonally up to the top left or bottom right by varying amounts to get the three peaks on top of each other. Then press enter.

Then, only when the three peaks are together (i.e. its roughly colour balanced) then you can clip all three colours by the same amount to the left of the main peak.

But you will really get so much more, and it will be so much easier, if you can remove the gradients first.

Good luck!

Cheers

Simon

[michael.h.f.wilkinson]

I shall have a read of your paper, thank you.

Thanks for the interest. The paper is quite strongly mathematical. If you need any explanations let me know.

[simon hicks]

Hi Michael,

I took your 'flattened' version of the image and gave it a bit of a histogram stretch to see what your algorithm had actually done. It gave the image below....

So it is clear that your technique has simply deleted all of the background data and deleted most of the fainter nebula data as well.

If you take Matthew's original image and flatten it properly and then stretch it you will get this....

Note that the second method has revealed all of the faint detail in the nebula that is actually in the original picture.

Cheers

Simon

[michael.h.f.wilkinson]

You are right of course, and I did not try optimize in any way, because I was curious what the newer method may achieve. Changing the area thresholds/noise suppression parameters needs tweaking, which I did not have time for. It is also possible to specify different properties for filtering, or developing different filtering rules, which could be optimized to deal with Poisson noise. I am working on the maths currently, and may have some better solutions shortly. Also, the current method does not make use of the correlations between the colour bands, which will be added later.

Of course, nothing replaces better acquisition.

[Matthew.Blake]

Thanks for the interest. The paper is quite strongly mathematical. If you need any explanations let me know.

I'm going to have to read it through a few times first and sit down with some imaging texts I think. It's been a while since I did any maths at that level and I'm a little rusty!

[michael.h.f.wilkinson]

I slightly more introductory paper on connected filters can be found here: http://www.cs.rug.nl/~michael/SPMfpage.html

[Matthew.Blake]

I slightly more introductory paper on connected filters can be found here: http://www.cs.rug.nl/~michael/SPMfpage.html

Ta muchly, I'll have a read through that then another look at the first one!

I tried the demo version of Gradient eXterminator on the image and got diddly squat improvement. Hurumph.

[simon hicks]

Hi Matthew,

I don't use GradientXterminator (what a great name!), I use the PixInsight version. But what I can say is that you want to set up GradientXterminator so that it is modelling the background away from the nebulosity....even the faint nebulosity on the right hand side of the image. I'm not sure how you do that in GradientXterminator, but that's what you want to be doing. If it includes the faint nebulosity in its estimation of the background then you are doomed from the start.

So maybe read a bit more about GradientXterminator or maybe someone here with more knowledge about it can help you tune the parameters.

Good luck!

[michael.h.f.wilkinson]

Many of these methods work based on a so-called rolling-ball (or top-hat) filter. I tried some that come with ImageJ, but was not impressed. There are other methods based on Fourier analysis, but they are not necessarily better. Quite fundamentally, if the contamination (whatever it is) has similar shape/size/brightness/colour properties to the object you want to image, you are in trouble.

[Matthew.Blake]

I've just had a look at the stacked version of some shots I took of M31 last night on the basis that it was in a different part of the sky. I get an almost identical gradient with an off centre bright spot. I'm going to see if I can get some flat frames and if that helps. Hmmm, can an LCD monitor be used as a light panel?

[simon hicks]

The bright spot just off centre is usually standard vignetting from the lens or mirror. Flats should remove that nicely. Yes, you can use an LCD as a flat frame image. Make sure the camera / telescope etc are focussed to infinity just as they were when you took the astro image.