RGB hell

[Datalord]

So, after messing up my flats and tinkering with that, I got what I believed to be a fair set of images for each channel. My lums are sweet, but the RGB part is total mayhem. As this is my first adventure into LRGB with a mono, I would appreciate anyone who knows better if you can produce something meaningful out of my data, so I can check whether I'm a processing idiot or my data is just not good enough.

These R, G and B files have been calibrated and stacked: https://www.dropbox.com/sh/ui5wlka2uvamwd6/AABpnBf67Mur2bE-A7Empnb6a?dl=0

This is the horror I managed to get out of them (Lum combined and further processed to hell):

 

[carastro]

I would happily take a look at them but I think they are Pixinsight format which I can't open.  Do you by any chance have a Tiff file?

I can't advise on Pixinsight processing but I can tell you if the data is any good.

Carole 

[Datalord]

5 hours ago, carastro said:

I would happily take a look at them but I think they are Pixinsight format which I can't open.  Do you by any chance have a Tiff file?

I can't advise on Pixinsight processing but I can tell you if the data is any good.

Carole 

Thanks. I just uploaded them as tiff.

[Oddsocks]

A very quick process in PI.

Problems with the flats in green and blue channels as dust bunnies remain in the source images, though very faint, needs a little tinkering with the settings in image calibration or your flat fielding method to sort that issue.

The first step working with the .xisf files is to dynamic crop and remove the registration overlaps then either DBE the individual colour channels to remove the light pollution gradients before channel combination, or channel combination first followed by DBE on the combined image. Either method is acceptable though for an image with complex gradients such as this you might find it easier to DBE the individual channels before RGB combination.

The steps performed to get to the stage below were:

Dynamic crop on individual channels using saved process (Dynamic crop on the worst channel, save the process and apply it to the other two channels).

DBE individual channels using saved process (DBE on the red channel, save the process and apply it to the other two channels to ensure the same sample points were used for each channel).

RGB Channel Combination.

SCNR to remove green artefacts.

Colour calibration using galaxy core for white reference and separate background sample with structure detection disabled.

Colour saturation.

Auto Histogram.

Histogram Transformation.

More can be achieved with the data you have. The colour calibration is not that good in this attempt, the Photometric Colour Calibration module would have been a better choice but I did not have the time to plate solve the image to get the astrometric data and just used the quicker basic colour calibration using white and background references. A good star mask would help to keep the background stars subdued, I find wide field images of M31 that have just had a basic stretch like this become too cluttered as the background stars are too prominent, a good star mask on the full size image would take me an hour or two to create but the results would have been better I think.

With a bright target like M31 I'm not sure luminance is worth the effort and the few times I have had an instrument with a wide enough FOV to image it I have found luminance did not bring much more to the result that just the RGB channels alone, perhaps it's time I had another go now that I have an IDAS in a spare slot of the filter wheel and can try using that instead of the normal clear luminance filter.

Don't be surprised to spend hours on a wide field image like this to get the best out of PI.

HTH.

 

 

[Datalord]

Oddsocks, thanks for your attempt. I pretty much did exactly what you have been doing, with the notable exception that I redid the DBE on each channel. Lesson learned. I still don't understand what went so wrong with my flats and I will never know, as I had to move the rig around to get a new OAG camera placed.

About the lum, I think that turned out quite ok for me. If anyone wants to use it for anything, I have uploaded my luminance. It is 124*30s, 32*180s 9*300 and 2*600. Drizzled 2x, DBE and Deconvoluted in 7044*5538.  I uploaded both a linear and a stretched+HDRMultiTransform version. 300MB each.

A screenshot of the linear STFd:

 

[carastro]

I have just finished doing the processing in Photoshop.  I found very bad gradients and had to do gradient exterminator on more than one occasion.  I also found the colour to be predominantly in the yellow, but that could be because I increased the colour to help improve the star colour.

I was just about to start a 2nd process whereby I did gradient exterminator on each of the coloured channels individually, but reading what has been written above, I think DBE is similar to GE, so as you have already reached that conclusion, I won't go on to do that second process.

Had great difficulty in extracting any blue in the image.

I wouldn't mind however having the link to the luminance to finish the image off in a tiff format please. 

Anyway, this is the result I came up with.

Carole

 

[carastro]

Now I have posted it up I can still see a further gradient at the bottom, but will leave it for now.

Carole 

[Datalord]

4 minutes ago, carastro said:

Now I have posted it up I can still see a further gradient at the bottom, but will leave it for now.

Carole 

I really appreciate the effort. Thank you. I have uploaded the lums in the same folder, and those are sort of ok to my eyes. Use at will.

[Oddsocks]

Had another quick process with the source .xisf RGB.

Slightly different process steps:

DynamicCrop of individual RGB.

DBE of individual RGB.

LinearTransformation of individual RGB.

ChannelCombination.

CurvesTransformation.

MorphologicalTransformation x 3 with Star Mask x 3

Plate solve at Astrometry. net to get arc sec/pixel and RA DEC.

PhotometricColorCalibration.

AutoHistogram.

ArcsinhStretch.

HistogramTransformation.

SNCR to remove remaining green artefacts.

Save As .jpg.

 

Still not quite there, a manually created star mask would have been better at keeping those tiny stars subdued, at this image scale they look too much like hot pixels .

William.