Finally got around to processing some Astrofest images. This one is just under 2.5 hours total integration using the QSI683, SN10, EQ8 combination.
I recently purchased Warren Keller's Inside PixInsight and have processed this image based on Warren's workflow and suggested settings. I've learn't a lot from the book so glad I bought it.
That looks pretty good, Peter. Did you use masked stretch?
I also have the inside PI book and found it very useful during processing .
Thanks David. I used HT for the Lum and MS for the RGB. I spent a lot of time comparing different stretching options (HT, MS, AH) for both the RGB and the Lum but came to the conclusion the differences were very subtle when compared with similar amount of stretch and black point. Warren recommended MS for RGB to preserve star colours so I went with that.
Looks really good Peter! It is quite a dim galaxy so you've captured it nicely.
My only suggestion processing wise is to use a star mask (or a stronger one) on the brighter stars as they have turned into a bit of a donut. This happens when doing an unmasked HDR the bright (potentially clipped) data doesn't scale very well.
The same look can be caused even without doing a HDR if during the Histogram Transformation you clip the top end (the brighter stars) to drag up the bottom and allow a better dynamic range.
Thanks for the feedback Colin. I did notice some problems with the brighter stars with a dark central spot but was still trying to work out which step caused that.
I was using Star Masks for various steps but I find it is a bit of an art so more practice required there!
A nice looking NGC6744, Peter I always get those hard looking star cores with MS. If they're coming from your colour data then that might be the cause.
Thanks for the feedback Colin. I did notice some problems with the brighter stars with a dark central spot but was still trying to work out which step caused that.
I was using Star Masks for various steps but I find it is a bit of an art so more practice required there!
After many many hours of playing with masking, I have found the best way is to make several different star masks and blend them together in PixelMath. I'll do different scales at 2,5,8 and sometimes an 11 when there are some super bright stars without too much nebulosity.
How do you decide how much to blend of each? Is there a way to look at the mask and the image and decide if the mask is going to work or do you have to apply the process and judge based on the result?
Oh that is what the Max(a,b,c) does, it takes the max value of the pixels. It does well at making the small stars (small scale of 2) and big stars big (big scale of 8+). There is no single scale that does well which is why I make big and small scale star masks.
Using the Max() function means you don't need to worry about the amount of blending from each layer.
Is there a way to look at the mask and the image and decide if the mask is going to work or do you have to apply the process and judge based on the result?
You can apply the star mask, invert it, then toggle Show Mask on and off to get an idea of how well covered your stars are.
You can apply the star mask, invert it, then toggle Show Mask on and off to get an idea of how well covered your stars are.
Thanks for that tip Rick. How do you determine how many/what size stars need to be masked? Also in linear processing steps should the mask be STF stretched when comparing to the image?
I personally try to mask as many stars as possible. The bigger ones are generally more susceptible to being destroyed via processing than the smaller dimmer ones. Masking the medium to bright ones is more important than the little ones.
I usually create a star mask for the linear processing and then another one after it has been stretched and then use a Max() function on both of those to pick up some of the stars that are easier in stretched and un-stretched images.
When actually stretching however, you don't want a mask active at the time. Try it and see what it does
Thanks for that tip Rick. How do you determine how many/what size stars need to be masked? Also in linear processing steps should the mask be STF stretched when comparing to the image?
Hi Peter,
The coverage I look for depends on what I'm trying to achieve. A mask to use as a decon ringing support only needs the very largest stars, so linear data with a small midtones stretch works well. A star mask to use for blurring chrominance (helpful to make star cores more colourful) only needs the large and middle sizes stars. A mask for star size reduction or star removal usually includes stars of all sizes. For this, Colin's suggestion of combining multiple masks with the PixelMath max() function is useful.
You can stretch a star mask and I do this occasionally especially when combining multiple masks, perhaps to strengthen the masking of smaller stars. You don't usually need to stretch a star mask just because it is based on or being applied to linear data.
A star mask is just an image so you can apply any PI process or script to it. I often use MorphologicalTransformation Erosion or Dilation to adjust a mask.
Applying HDRMT with a very heavy hand (say 4 or more iterations) is often a good way to pre-process a copy of an image before running StarMask, especially if there are stars buried in nebulosity.
Hi Peter,
I hope you don't mind?
I had a go using Fitswork4 to reduce star sizes &
I also played around in LAB mode to get some more colour
& change the colours to show the red & yellow parts better in the galaxy.
Do you think this looks better?
I always get those hard looking star cores with MS. If they're coming from your colour data then that might be the cause.
