Color Noise Removal

[GeneralT001]

Hi,

I am really, really new to Pixinsight and am doing my best to come to grips with it. Its a great program and I am dedicated to learning it. I was hoping I could get some help with the following issue. I am trying to process some M13 data but cannot get rid of some color (orange) noise. I am at the point where I have done Dynamic Crop/DBE/Background Neutralization/Color Calibration/Histogram Stretch/Luminance Mask +Multi Scale Linear Transform (to reduce noise). I have attached a picture of what I see when zooming in.

This was also dithered.

I have also provided a link to the >Fit file if anyone wants to have a crack at it.

Is it possible to get rid of that noise ( data was taken on a full moon with only an hour of data). I have had this same color issue with 5 hours of data so was hoping to figure out a way to get rid of it before committing to longer data acquisition. If it is possible can you explain (laymans terms) how to do it or point me towards a tutorial. I have looked at a lot of tuts but cannot find what I need (maybe it doesn't exist)?

Many Thanks

https://www.dropbox.com/s/lb2cudedbanda6w/M13_79subs.fit?dl=0

[jkmorse]

From the sounds of it, you are shooting color images and not monochrome with filters.  One thing to think about is using channel extraction to separate your images into separate RGB images that you could then try tackling with MureDenoise which is a superb noise reduction script but it only works with monochrome images. 

I have also written up a number of other techniques that may help in my PixInsight workbook that I freely share with others (if you are interested in the whole workbook, just drop me a line at jkmorse57@gmail.com and I will be happy to send it along).  Here are some ideas:

i.   Background Repair
There are times when, in the midst of processing, your background gets out of whack.  That can take any number of forms, including color blotches, uneven noise distribution, and a host of other problems that can seriously compromise the look of your image.  Here are a few techniques I use to help solve some of these issues, often in combination.  In each of these, it is critical to build a good lightness mask so that the stars and target are protected and you are only working on the background (and don’t forget to invert that mask so it’s the background that is white and everything else dark).  Also, I generally only apply these techniques after an image has been stretched and typically only apply them as a last resort since they are not true to the data (the best solution by far is simply driving up your signal to noise ratio which is a much better way to handle the problem):
•   Color Correction
This is an easy first step and one I apply to almost all my images.  Often, despite your best efforts with DBE, you still end up with a background that has blotchy colors rather than a uniform grey.  To fix, create a highlight mask, covering everything but the background.  Invert so everything but the background is protected.  Open the CurvesTransformation tool and select the Saturation button (the “S” button).  Create a smooth curve by pulling DOWN at the center point.  Apply to background.  Can be done more than once to eliminate any residual colors.  Point is to arrive at a neutral grey, even if it remains blotchy.  We will fix that in the next steps 
•   Background Smoothing
Step 1: Convolution – Ensure that the image still has the mask protecting everything but the background.  Open the Convolution tool and set the Std Dev to around 5 and apply to the background.  You may need to apply this more than once.  The background will now be smoother, but may still be a bit blotchy so on to step 2.
Step 2: Morphological Transform – With the mask still in place, open the Morphological Transformation tool and insert the following settings:
•   Set Operator to Morphological Selection
•   Set Interlacing to 11
•   Leave Iterations to 1
•   Set amount to around 0.35
•   Set Selection to 0.50
•   Set Size to 25 (the highest setting)
•   Leave Way at 1 of 1
•   Select Circle (the second box in the second row)
•   Apply to the background (more than once if necessary)
At this point your background should be dramatically better.  In fact I sometimes find it so good that I need to reintroduce a bit of noise using the Noise Generator tool (using the Uniform Distribution setting).  Credit for this great technique goes to Mike Smythers at this post: http://pixinsight.com/forum/index.php?topic=8591.0
Also, check out this post for some other great processes, especially related to masks, developed by Mike.  If you find these as helpful as I do, send him a thank you:
http://pixinsight.com/forum/index.php?topic=8618.msg55913#msg55913


Hope that helps,

Best,

Jim

[aworonow]

Jim,
  I see that you set interlacing to11. I've never seen any setting other than 1 mentioned before. Could you explain a) that INTERLACING is, and b) why 11 and not, 10? (that is, how does one know when the best value is set?)


Thanks much, Alex

[msmythers]

Jim
I'll take the question about the interlacing.


Alex
I don't have a mathematical reason for the interlacing number. I experimented on a lot of images. I varied all of the setting in MorphologicalTransformation until I decided that these numbers gave satisfactory results over a large range of different types of images. The key to this method working well is a good mask of all the non-background objects in the image.

Here is another posting where it might be a little easier to see and understand what I think is happening with the Interlacing setting.
http://pixinsight.com/forum/index.php?topic=9149.msg58756#msg58756



Mike

[GeneralT001]

Thanks jkmorse,

I tried your suggestions.

The convolution made things quite fuzzy, so I didn't use that in the end.

The Morphological Transformation did have a very positive impact. So much so, that I don't think I can expect much more from the image given the conditions it was captured under - many thanks!!

One other question, if I may, is it common to mask the background when applying saturation? I would guess it is as you may only want to bring the colors of the stars up? I would think if you applied saturation without a mask you may make any color noise in the background worse?


Thanks

[aworonow]

Mike,
Thanks much. I get it. Tried out a bunch of settings and I see the general trend.

Alex

[jkmorse]

General,

Yep, any time you are bringing out detail in the target, whether that be saturation, sharpening, structure enhancement, etc, you should be masking the background, both in the linear stage and after stretching.  This gets more complicated with lots of nebulosity, but it is still important to protect the areas you are not trying to enhance.  And with some techniques, such as sharpening, you really want to build a combined mask that protects both the background and the stars and isolate only the target.  Best way to do that is with pixelmath. 

Masks are your friend, learn them, use them.

And Mike, thanks for handling.  I would have handed off to you anyway since you get all the credit for that great technique.

Best,

Jim

[GeneralT001]

General,

Yep, any time you are bringing out detail in the target, whether that be saturation, sharpening, structure enhancement, etc, you should be masking the background, both in the linear stage and after stretching.  This gets more complicated with lots of nebulosity, but it is still important to protect the areas you are not trying to enhance.  And with some techniques, such as sharpening, you really want to build a combined mask that protects both the background and the stars and isolate only the target.  Best way to do that is with pixelmath. 

Masks are your friend, learn them, use them.

And Mike, thanks for handling.  I would have handed off to you anyway since you get all the credit for that great technique.

Best,

Jim

Thanks. Appreciate the confirmation!