Trouble with Clarity in PixInsight

[SpaceBob]

I am brand-new to PixInsight and have followed a few YouTube videos trying to find my way, and don't get me wrong, these photos are better than what I used to be able to do but I'm also seeing some glaring problems with them. I've attached my "results": M81 and M82, the Flaming Star Nebula (I think), Leo's Triplets, and M51. I used about 20 subs each at 5 minutes with the proper amount of dark frames (but no flats or biases). Stacked in DSS; in PixInsight I used STF, ABE, Background Neutralization, Color Calibration, Histogram Transformation, Range Selection, LRGBCombo, Curves Transformation, SCNR, LHE more or less in that order. The problems I'm seeing are excessive noise, lack of proper colors and detail in the structures. Is this something to do with how I capture these objects or how I process them? Any help would be appreciated.

Location: Florida, elevation ~100 ft, almost as dark as a dark sky site

10" Celestron Newtonian, Advanced GT mount with autoguider

Nikon D5500 DSLR, 2" GSO Coma Corrector

[SteveA]

With what sounds like a very good location (most of us on here would probably kill for a near dark sky site), I think I agree with you that you should be able to produce better images...(sorry if that sounds harsh).

One thing that jumps out from these images is that you appear to have a guiding issue. Very egg shaped stars are pretty much indicative of that and this may well be the cause of the issues that you have identified yourself. A lack of flats may also contribute. I would address the guiding issue first and maybe even something as fundamental as a poor polar alignment may be to blame.

Hope that helps...though I'm sure others will also have some comments.

Steve

[Oddsocks]

Without being able to see the source raw frames it’s going to be almost impossible to say what is wrong from looking at an end result. If I had to guess I would say the problem was at the calibration stage, possibly the missing bias and flat frames or incorrect setting and handling of pedestal values, or simply too few frames in poor seeing conditions.

The misty nebulosity surrounding the bright stars in the Leo triplet and M51 images hints at dew on the optics, or high atmospheric cloud and ice, if thats the case you should see that in the single raw frames for those objects too, if the raw frames don’t show the effect then maybe this is the result of incorrect application of ABE.

In your processing sequence you are using ABE which was replaced by DBE sometime ago. ABE still has some use and is still included in PI but for your images DBE is the appropriate choice because it better handles noise assessment in the source image before flattening.

When using DBE always inspect the rejection maps and subtraction/division masks before deleting them, use STF on the masks to visualise better, the masks should show a smooth even gradient, either a vignetted pattern, linear pattern or a combination of the two, but it should not show a defined structure, if it does then it means data is being removed and lost and the initial settings or sample point placements used in the DBE tool are not optimal.

DBE and ABE are no substitute though for properly calibrated source data. DBE is optimised to remove vignetting by division or gradients by subtraction but won’t handle both simultaneously with the best of results so get those flat frames included in the calibration routine and use DBE only to remove the residual gradients while leaving as much of the original data intact.

For DSLR images better results can be obtained by dropping the dark frames entirely and using a master bias together with a bad pixel map, mount dithering and flats. For DSLR, bias frames are probably more important than darks however you should still see reasonable results. Where you do include darks you need lots of them otherwise you just add noise.

When I occasionally used a Nikon D90 I found I needed around 60 or so darks and bias frames and at least 40 image frames to keep noise under control but then changed to a bias master of 80 frames, 40 flats, no darks, mount dithering and bad pixel map and saw improved results.

There is no need to use DSS before PI. PI can take the raw Nikon frames, calibrate and stack, all in the one program, and it doesn’t supress the colour information the way that DSS tends to do, so do try a calibration and stack in PI and see if that is different.

Always compare a calibrated image before processing in PI against the outcome of each processing tool. I always carry out post processing in PI on a clone of the calibrated, registered and combined image and keep the original open on the desk top, comparing back to the original unaltered image after each processing step on the clone, that way you can see early on where a particular tool or processing sequence begins to go awry and the image degrades, simply using the “undo” “redo” buttons on a post processing tool doesn’t really help you to keep track of the changes over a sequence of tools but keeping an unaltered calibrated original open on the desktop you will have something to compare against.

HTH

 

[SpaceBob]

Thanks for the info guys. I'll admit that the scope is a little much for the mount but even then, PHD2 usually reports a mean error of .3 or less, with most of the error attributed to unaccounted backlash in the DEC motor, so that's probably why I'm getting egg-shaped stars. I was told an error of 1 or less was fantastic but maybe not. The seeing lately has been terrible where I am so that's probably a factor. I've also been lazy when it comes to calibration frames but now I know they're necessary. I'm going to recapture these images next time I go out imaging and I'll try to post back with results in a week or so hopefully. I was also told dark frames are the most important out of the calibration frames so I'll experiment without the dark frames and with the flats and biases.

Again, thanks for the advice!

[SpaceBob]

I know I'm quite a bit late in following up to my original post but I redid my M51 photos, this time using flats, biases, darks, and of course, lights. The first one is a sub (it was quite windy) at 3 minutes, the second after stacking in PixInsight, and then the third is after my workflow, which is the same as before but with DBE this time instead of ABE. I used the same equipment in capturing these photos too, as well as being at a fairly darker site this time around. I know my colors are a bit off and having no wind would help a lot, but anything else I could do to improve the quality? On a side note, I also tried using Deconvolution which added quite a bit of detail but destroyed the stars no matter how many times I redid the star masks and played with the settings, so I eventually gave up on it. Thanks for the help!

[wimvb]

PixInsight is a power tool. As such it must be handled with care, or it will grind away at your data. The images look a lot better already than before. But the final image still looks overprocessed to me.

If you're willing, post the stacked but unprocessed image file in xisf or fits format, and let us have a look at it. (Note, xisf files tend to be large.)

[Oddsocks]

8 hours ago, SpaceBob said:

I also tried using Deconvolution which added quite a bit of detail but destroyed the stars no matter how many times I redid the star masks and played with the settings, so I eventually gave up on it.

 

As Wim states PI needs careful handling and deconvolution is one of the trickiest to get right.

Pixel peeping at your stacked results I can still see lots of hot pixels and suspect you didn't use any pixel rejection in the Image Integration module. Use Sigma Clipping or Winsorized Sigma Clipping in the pixel rejection (1) tab of the image Integration Module if using the standalone Integration Module or set Sigma Clipping or Winsorized Sigma Clipping as the rejection algorithm in the Image Integration tab of the Batch Processing Script if you used that.

The tracking problems have robbed much of the detail from the image and in addition there is a noticeable difference in star shape across the image, suspect that there is some tilt in the optical path. Maybe this is collimation or the camera is not orthogonal to the flattener. Some 'T' rings are prone to flop and don't hold the camera tightly to the flattener and that is where I would look first followed by the fit of the flattener into the focuser draw tube.

When using a large heavy OTA guiding is more difficult and wind doesn't help but in your images the trailing is all in the same direction and this may be backlash in the RA drive or over aggressive settings in the guiding module, try unbalancing the OTA slightly so that the OTA is slightly east-side heavy and adjusting the guiding to be less aggressive. You often see better guiding results by not trying to correct out all the guide error in a single step but correcting just a part of the error so that the OTA doesn't continually oscillate between corrections. I don't know your mount but if it has permanent PEC you may see an improvement in guiding when working close to the weight limits if PEC is employed and give the guiding less work to do.

Wim will attempt a reprocess of your image if you attach a stacked XISF here or a link to a Dropbox or Google Drive shared folder.

Just for fun I took your final posted 8bit jpeg back into PI and applied several small separate masked iterations of the Deconvolution Module for Parametric PSF and Motion Blur, selecting an angle to match the trailing and then used the ACDNR module to remove some of the noise, finally a little cosmetic correction in Photoshop. Image posted below.

It wasn't worth spending too much time on the low resolution forum image but it hints that more is possible with your data. 

Overall though you must be pleased that things are going in the right direction!

William.

 

 

 

[SpaceBob]

I've attached a Google Drive link for the .xisf stacked image; thanks for doing this wimvb. If you're willing, could you document what you did differently than my workflow so I know what to improve on?

As for pixel rejection, I used sigma clipping under Pixel Rejection (1) and I just recently figured out how to mechanically correct backlash in my DEC motor housing; the RA drive has little to no backlash. I guess I need to take another look at how my OTA is balanced. I'll admit my aggressiveness settings were high; I'll take a look at those too when I go out next. 

I can see where Deconvolution helped to define the structure; thanks for showing me that.

https://drive.google.com/open?id=1vefLzXsL09-mloVOT2U_KFh4nQ3R7OxU

[wimvb]

I'll have a look. But the next few nights are supposed to be clear, so I'll be imaging. It may be a while before I'll report back with results.

[SpaceBob]

Well, in regards to Deconvolution, I figured out that I had the deconvolution mask inverted so that's why it was impossible for me to get that to work. I'll post back in a bit with my new results but they're looking beautiful.

[SpaceBob]

So the first one is my final result. I have no idea what those dark squares are around the three bright stars on the left, I think that's something leftover from Deconvolution. I tried not to over-process it but I'm not entirely sure what would be over processing and what wouldn't be (any advice?). The second one is the stacked image before processing, i.e. before Deconvolution. The third one is after Deconvolution but before anything else. I think I'm starting to see where having a CCD would really help in cutting down on noise and on bringing out more color. Could I try shorter exposures and more subs with my DSLR or no?

[wimvb]

Processed in pixinsight. I didn't crop the image and left the bright area in the bottom as it was.

(click on image for larger version)

Main process steps:

• DBE
• Colour calibration
• Deconvolution to reduce star trails
• Stretch and colour saturation
• Noise reduction
• Star reduction to reduce star trails
• Resample

m51_spacebob.xpsm