Teasing out the detail

[Ivor]

I have been working on an image of M81 & M82 so far I've completed the preprocessing steps, Dynamiccrop, DBE, BackgroundNeutraization, photometricColorCalibration and SCNR. I'm not sure how best to proceed in PI, if I have play with Histogramtransformation I see with each increasement levels of detail I'd like to keep, should I be creating masks at each increament before applying another stretch? I've not found a tutorial on approach so I'd be grateful for any tips.

 

 

M82_int_light_crop_DBE_BN_PCC_SCNR.xisf

Edited January 29, 2023 by Ivor
Updated file

[Ivor]

  Further research and I have discovered the power of HDRMultiscaletransform, and I have successfully got my image further. I'd appreciate some advice on how to stop the effect around the star in the middle of M81, I’m guessing I’ve been too aggressive somewhere but unsure where. I also tried to bring out the dark lanes in the core of the galaxy using the DarkStructureEnhance script but that didn’t really work either is there another way I can investigate?

 

[Budgie1]

Is the image file you've uploaded the stacked image, or have you already performed DBE, BackgroundExtraction, PhotometricColourCalibration and SCNR on the image before you uploaded it?

I ask because it seems to be lacking some colour, mainly the red is very weak and I would expect more red to show from the centre of M82. If you have worked on the image before upload, could you upload the raw stacked image?

The use of Curves with RGB selected and fine manipulation of the top right of the curve can make the centre of M81 more uniform, but the centre is very bright in your image and there's very little detail in there. If you remove the stars using StarNet2, you could create masks to isolate M81 & M82 to allow you to work them separately. This may help you control each galaxy better,

[Ivor]

You are right this image has been though these process, I was unsure why I'd lost so much color but trying to address one issue at a time. I've attached a version of the stacked cropped image pre DBE. Thanks for the StarNet2 tip I'll investigate than tonight.

 

M82_int_dizzle_crop.xisf

Edited January 31, 2023 by Ivor

[Aramcheck]

Hi @Ivor

Nice data! I couldn't resist having a quick play. Workflow as follows:-

• DBE
• ImageSolver (script)
• SpectrophotometricColorCalibration
• BlurXterminator (to Luminance only) (altenatively do deconvolution after extracting Luminance)
• RGBWorkingSpace (set values to 1)
• Extract Luminance
• rename original RGB

with the Luminance image I then did:-

• 3 slight stretches with HistogramTransformation, with the Midtone stretched to 25%
• StarXterminator (to generate stars only & starless Luminance images)
• 3 further slight streches to the starless image
• Applied a small gradient mask to the core of M81 (using Hartmut V. Bornemann's free GAME script); invert & a further slight stretch
• Invert the mask & apply a CurvesTransformation to brighten the core slightly.
• On the stars only luminance image, I did four slight streches with HistogramTransformation, with just the Midtone stretched to 25%
• Used PixelMath to recombine the Luminance stars / starless images using: ~((~starless)*(~stars)) and rename "Lum"
• Create a clone of the resulting Luminance image, then apply a mask around M81 and apply HDRMultiscaleTransform and then LocalHistogramTransformation. (I create a preview of just the galaxy, make several copies and then try different parameters until I'm happy with the result & then apply those settings to the main image). Then repeat with a mask around M82... finally remove mask and rename "Lum_hdrmt"
• Use PixelMath to combine the "Lum_hdrmt" and "Lum" images - in this case with 0.5*Lum + 0.5*Lum_hdrmt and rename "L50"

On the RGB image I then did:-

• Repaired HSV Separation (script)
• ChannelCombination using the output from the above but with the "Unrepaired_V" file & rename "RGB_Repair"
• Clone "RGB_Repair" and apply MaskedStretch & rename "RGB_MS" (I usually also similarly clone / run ArcsinhStretch and clone / run the EZ SoftStretch script & then use Pixelmath to create a blend... in this case I just chose the MaskedStretch result)
• On RGB_MS i then ran ChannelExtraction to create HSI components. Delete the "I" image
• use ChannelCombination with HSI selected, using the H & S images created from the RGB_MS file and the "L50" as the "I" component. Rename the result as "HSI"
• Extract the luminance from the HSI file and using ScreenTransferFunction / HistogramTransformation create a high contrast image to use as a mask to adjust the colour saturation (rename: Sat_mask)
• Apply "SatMask" to the HSI image & use Curvestransformation to increase saturation. (Two adjustments were made)
• Apply SCNR with a small reduction to the Green channel (reduced to 0.9 x original)
• Apply NoiseXterminator (if using other noise reduction techniques then I would instead do some noise reduction when still linear after deconvolution and then after recombining the stretched images)
• To reduce the size of Stars I then ran StarExterminator again on a clone & renamed the results "Starless", before running Bill Blanshan's star reduction script... and a final tweak with HistogramTransformation

Thanks for sharing the data!
Cheers
Ivor

[Ivor]

I have 3 mins before I head off to work however wow, thanks so much for the detailed process breakdown I'll digest this and come back with a few questions.

[teoria_del_big_bang]

22 hours ago, Aramcheck said:

Hi @Ivor

Nice data! I couldn't resist having a quick play. Workflow as follows:-

• DBE
• ImageSolver (script)
• SpectrophotometricColorCalibration
• BlurXterminator (to Luminance only) (altenatively do deconvolution after extracting Luminance)
• RGBWorkingSpace (set values to 1)
• Extract Luminance
• rename original RGB

with the Luminance image I then did:-

• 3 slight stretches with HistogramTransformation, with the Midtone stretched to 25%
• StarXterminator (to generate stars only & starless Luminance images)
• 3 further slight streches to the starless image
• Applied a small gradient mask to the core of M81 (using Hartmut V. Bornemann's free GAME script); invert & a further slight stretch
• Invert the mask & apply a CurvesTransformation to brighten the core slightly.
• On the stars only luminance image, I did four slight streches with HistogramTransformation, with just the Midtone stretched to 25%
• Used PixelMath to recombine the Luminance stars / starless images using: ~((~starless)*(~stars)) and rename "Lum"
• Create a clone of the resulting Luminance image, then apply a mask around M81 and apply HDRMultiscaleTransform and then LocalHistogramTransformation. (I create a preview of just the galaxy, make several copies and then try different parameters until I'm happy with the result & then apply those settings to the main image). Then repeat with a mask around M82... finally remove mask and rename "Lum_hdrmt"
• Use PixelMath to combine the "Lum_hdrmt" and "Lum" images - in this case with 0.5*Lum + 0.5*Lum_hdrmt and rename "L50"

On the RGB image I then did:-

• Repaired HSV Separation (script)
• ChannelCombination using the output from the above but with the "Unrepaired_V" file & rename "RGB_Repair"
• Clone "RGB_Repair" and apply MaskedStretch & rename "RGB_MS" (I usually also similarly clone / run ArcsinhStretch and clone / run the EZ SoftStretch script & then use Pixelmath to create a blend... in this case I just chose the MaskedStretch result)
• On RGB_MS i then ran ChannelExtraction to create HSI components. Delete the "I" image
• use ChannelCombination with HSI selected, using the H & S images created from the RGB_MS file and the "L50" as the "I" component. Rename the result as "HSI"
• Extract the luminance from the HSI file and using ScreenTransferFunction / HistogramTransformation create a high contrast image to use as a mask to adjust the colour saturation (rename: Sat_mask)
• Apply "SatMask" to the HSI image & use Curvestransformation to increase saturation. (Two adjustments were made)
• Apply SCNR with a small reduction to the Green channel (reduced to 0.9 x original)
• Apply NoiseXterminator (if using other noise reduction techniques then I would instead do some noise reduction when still linear after deconvolution and then after recombining the stretched images)
• To reduce the size of Stars I then ran StarExterminator again on a clone & renamed the results "Starless", before running Bill Blanshan's star reduction script... and a final tweak with HistogramTransformation

Thanks for sharing the data!
Cheers
Ivor

 

Great processing and thanks for sharing your detailed workflow, it helps more people than just the OP

Thanks
Steve

[Aramcheck]

I should have pointed out that processing the Luminance separately to the RGB, allows you to control the eventual colours better. In Pixinsight's normalised scale where Black = 0 and White = 1, the apparent colour in any pixel will be lost as the value increases, so to retain the colour towards the galaxy core you want to keep as much of it at values of 0.6 - 0.8, or at least limit the region where the pixels are close to saturation. Applying a gradient mask part-way through stretching the Luminance helps to prevent brightening the core too much, but will then require inverting the mask & applying a softer stretch so that it still appears to get brighter as you get closer to the centre. (It's not something you'd do all the time, but it is an option you can use)

It used to be that deconvolution worked best on Luminance only, but in the workflow above I made a mistake applying BlurXterminator only to the Luminance. Watching Russ Croman's youtube interview with Adam Block, it seems that this new tool works best applied to the RGB prior to splitting off the Luminance.

Warren A. Keller's book "Inside Pixinsight" is a good primer for the Pixinsight, even though it is now quite out of date. I'd also recommend anything on youtube by Adam Block & if you like his approach to PI/teaching method then his Fasttrack and Fundamentals video tutorials are worth buying:- https://www.adamblockstudios.com/

Cheers
Ivor

[Ivor]

Thanks for the additional advise, I've not purchase any of Xterminator tools yet, so I'm adjusting the workflow and branching off a few directions to see what works best. I'm adapting your workflow and using Starnet2 instead I found I got the best results with a standard STF first and then applying it. The edges of the larger stars are bit rugged so I'm researshing now best to clean it up, cleaning the starless image should be straightforward with clonestamp I'm unsure if I'm being too fussy with the starmask though. Thinking to the next step I'm investigating whether deconvolution to the starless image is the better option over MultiscaleLinearTransform.

 

[Ivor]

Had a chance to go through the complete workflow, aspect are better than my first attempt but I need to take this through a few more times. The background is clipped too much however I feel the detail in the galaxies has come out better. While I wait for the data for my next target to collect I'll play some more. @Aramcheck From one Ivor to another thanks for all your help.

 

[Aramcheck]

Before I bought the XT tools I used to largely follow Warren A. Keller's workflow in the "Inside Pixinsight" book. i.e. using Deconvolution to the Luminance (after DBE) followed by MLT for noise reduction & then Histogram Transformation. Once the non-linear Lum & RGB are combined (using LRGBCombination or switching the I in a HSI extract) I used to do a second MLT to effectively sharpen the image & then UnSharpMask.

One thing to look into is Pixinsight's HistoryExplorer. You can use it to show all the processes which have been applied to an image & go back to an earlier point in the processing. You can also drag any process which is in the history to the desktop, so if you want to make a change to one process, but want a quick method of re-applying the subsequent steps, you just have to drag the subsequent processes to the desktop, go back to the earlier stage & make your changes & then re-apply the saved processes by dragging them in turn from the desktop back to the image. Hope that makes sense - if not someone no doubt will have demonstrated examples on youtube. (I'd post some screen grabs, but my pc is currently tied up with comet data for the next few days!)

Cheers
Ivor