Heart of the Heart - First go with PI

[CraigT82]

Have downloaded Pixinsight and have been beavering my way though the light vortex tutorials.  There is so much to learn with pixinsight but I'm very happy with the way it has handled the calibrating, registering and stacking, certainly seems to have done a better job than APP did with the bin2x2 data of the same target from last week.  Well saying that there is a couple of warm columns at the top right which have not calibrated out.

I tried the multiscale linear transform tool in PI for noise reduction (using an inverted mask to protect signal) but it pretty much destroyed the image no matter what scale I used. I think I'm definitely doing something wrong there

This is 20x1200s Ha (Baader 7nm) with a 200p f/6 newt and 383L at bin1x1 (0.90"pp). 

Stacked in PI and with a masked stretch, then into Gimp for noise reduction then a bit of curves and contrast tweaking.

Edited March 3, 2021 by CraigT82

[TerryMcK]

Hi Craig. Try the new weighted batch preprocessing script in Script>Batch Processing>WeightedBatchPreProcessing

If using CMOS add your lights, darks and flats - use your normal subs for CCD (I think that is just lights and flats?). Click on the control panel tab and experiment with clicking a few of the parameters. Set your output directory and click Run and you will get pretty good results. You can put all subs taken with all three filters in at once and they will be sorted into Ha, SII, OIII etc dependant upon the FITS header.

There are 3 free tutorial videos from Warren Keller here if you want his explanation on how to tweak things.    

https://www.mastersofpixinsight.com/wbppv2

 

[TerryMcK]

Use MLT before stretching or as it is called histogram transformation. Use TGVDenoise after stretching.

[vlaiv]

Reported size of the image is 6708 × 5056 and 383L has less pixels than that.

You also mention 2x2 bin.

Did you by any chance use drizzle? Image is grossly over sampled and you should in fact aim for lower sampling rate not higher.

This is 1:1 or 100% zoom

but it should have something like quarter of current size - like so when viewed 1:1

 

[CraigT82]

30 minutes ago, vlaiv said:

Did you by any chance use drizzle? Image is grossly over sampled and you should in fact aim for lower sampling rate not higher

This image was captured at bin1x1 but part of the tutorial I was following was to fo drizzle integration at scale 2 so I did that, I guess you're meant to scale it back down afterward? Or just not bother with the drizzle part? 

[vlaiv]

2 minutes ago, CraigT82 said:

This image was captured at bin1x1 but part of the tutorial I was following was to fo drizzle integration at scale 2 so I did that, I guess you're meant to scale it back down afterward? Or just not bother with the drizzle part? 

Don't bother with drizzle part at all.

Drizzle is (questionably) useful when you are quite a bit under sampled. It tries to recover detail otherwise present but not captured due to low sampling rate. I maintain that it is not working in amateur setups as original algorithm was developed for Hubble and had precise pointing as requirement (like exact 1/3 pixel aligned dithers). Amateur setups can't do that and any dither is random in nature. I'm not convinced that random dithers in combination with drizzle algorithm in fact have any benefit of original algorithm.

In fact, if you are at 0.9"/px - I would advise you to do the opposite - use x2 binning instead. It is very likely, and judging by the image above that you are over sampled at 0.9"/px and that just wastes SNR.