Learning Monochrome Data Processing - Forums

Hello all,

I made the switch recently from OSC to mono and have been learning and creating a new processing workflow. This is what I have so far:

WBPP
Dynamic Crop
DBE
Linear Fit
BlurX/NoiseX/StarX
Stretch
Channel combination (process or pixelmath)
Curves
Lum mask and sharpen
Recombine stars and starlets

Where does color callibration or SPCC fall here? A lot of workflow examples I came across do not use it. Also, I understand BlurX works better on color image but I wont be creating mine until its been stretched.

Is there any other steps I could add here? Overall it seems fairly straightforward and I can have some fun with channel combinationations through Foraxx or some other methods. Hopefully I can collect the last bits of Sii tonight and can begin working on my final image this week.

Thanks in advance

If I don't have RGB data to go with SHO data I'll use the correctmagentastars script to make the star colors better.

As for blurx. I'll use channel combination right before stretching. BlurX, then I'll separate the channels again. I'm pretty new to narrowband processing myself, so take that for what it's worth.

I would definitely recommend getting some short exposure RGB images along with whatever DSO you image. That was something that I didn't anticipate when switching to mono.

You'll want to do Image Solver (astrometry) then SPCC. I've been doing that on my stars_only image but I'm actually not sure where its best done in the process.

Also, definitely experiment with Drizzle 1.5x or 2x in your pre-processing. According to some sources, this really works well with Blurx

This would be mostly for stars correct. I've seen posts with 30 second rgb for an hour each which wouldn't be too difficult to do this week as weather is clearing up.

Yeah still not sure what to do with SPCC. I could experiment with and without and see what happens.

Haven't tried drizzling before as my camera sensor and focal length give me good sampled data. But again, always something to try out and see what happens. Thank you!

If I don't have RGB data to go with SHO data I'll use the correctmagentastars script to make the star colors better.

As for blurx. I'll use channel combination right before stretching. BlurX, then I'll separate the channels again. I'm pretty new to narrowband processing myself, so take that for what it's worth.

Interesting, I can try it out both ways and see what happens. Thank you

I'm thinking about making the same change… OSC to mono, but not to use narrowband. I'm trying for a "natural" look to my images. I've processed a number of mono RGB or LRGB data sets, and I really think this approach gives better color than OSC. This is the process I use with mono. I try to do blurx as soon as possible so that images have not been changed much by prior processing.

1. Star Align all
2. Crop all
3. DBE, ABE or GraXpert on R, G, and B. I have found instances where GraXpert does a better job, but usually DBE is my goto.
4. Blurxterminator (I do this as soon as possible) on R, G, B
5. Noisexterminator on R, G, B… usually at about 80%
6. For R, G, and B do linear fit against the image with the lowest mean
7. Combine RGB
8. Image solve RGB combined if needed
9. SPCC… rarely fails
10. SCNR at 80%
11. Stretch, usually with GHS; followed by curves if needed for color and/or lightness. Sometimes use maskedstretch or EZsoftstretch or STF/HT
12. Luminance… same general process as R, G and B
13. Apply Local Histogram Equalization and or Unsharpmask to improve contrast of luminance. May use Advsharp script to sharpen.
13. Combine L and RGB after a slight blur of RGB… can remove a multitude of issues in RGB
14. Often use one of Bill Blanshan's star reduction processes
15. Photoshop editing if needed… usually not much

I would certainly appreciate feedback on this process.

George Hatfield:
I'm thinking about making the same change... OSC to mono, but not to use narrowband. I'm trying for a "natural" look to my images. I've processed a number of mono RGB or LRGB data sets, and I really think this approach gives better color than OSC. This is the process I use with mono. I try to do blurx as soon as possible so that images have not been changed much by prior processing.

1. Align all
2. Crop all
3. DBE, ABE or GraXpert on R, G, and B. I have found instances where GraXpert does a better job, but usually DBE is my goto.
4. Blurxterminator (I do this as soon as possible) on R, G, B
5. Noisexterminator on R, G, B... usually at about 80%
6. For R, G, and B do linear fit against the image with the lowest mean
7. Combine RGB
8. Image solve RGB combined if needed
9. SPCC... rarely fails
10. SCNR at 80%
11. Stretch, usually with GHS; followed by curves if needed for color and/or lightness. Sometimes use maskedstretch or EZsoftstretch or STF/HT
12. Luminance... same general process as R, G and B
13. Apply Local Histogram Equalization and or Unsharpmask to improve contrast of luminance. May use Advsharp script to sharpen.
13. Combine L and RGB after a slight blur of RGB... can remove a multitude of issues in RGB
14. Often use one of Bill Blanshan's star reduction processes
15. Photoshop editing if needed... usually not much

I would certainly appreciate feedback on this process.

For my clarification, Align is the Star Align process? This is to make sure the three masters are all aligned (I've only used this during mosaics but I understand it conceptually).

Our workflows do seem very similar outside of when to combine channels. For Linear Fit I've been seeing a lot of folks use the Sii channel as the reference frame but I guess this is dependent on the data?

The only thing I would double check is if Blanshan's method involves deconvolution. My understanding of BlurX is no other form of deconvolution is needed. I cant remember this detail from his script.

There is also this video which shows how to obtain "natural" colors with narrowband data.

https://youtu.be/_v9dnoZC1wI

Robert: Thanks for the link on "natural" narrowband images. Here are a couple of videos that may answer your questions about Blurx and Bill's star reduction methods. I don't think the latter involves using deconvolution. By the way, Russell Croman does a great job of explaining deconvultion in this video.

Good video where Bill describes his star reduction methods:  https://www.lightvortexastronomy.com/tutorials.html

Russell Croman video on Blurxterminator, etc:  https://www.youtube.com/watch?v=5_Dg7bYu-E8&t=5411s

Thank you so much George

Robert Khoury:
There is also this video which shows how to obtain "natural" colors with narrowband data.

interesting video. I've spent more of my time processing in SHO and not attempting a more natural image color like RGB. But now...thinking I'll have to give it go

George Hatfield:
I'm thinking about making the same change... OSC to mono, but not to use narrowband. I'm trying for a "natural" look to my images. I've processed a number of mono RGB or LRGB data sets, and I really think this approach gives better color than OSC. This is the process I use with mono. I try to do blurx as soon as possible so that images have not been changed much by prior processing.

1. Star Align all
2. Crop all
3. DBE, ABE or GraXpert on R, G, and B. I have found instances where GraXpert does a better job, but usually DBE is my goto.
4. Blurxterminator (I do this as soon as possible) on R, G, B
5. Noisexterminator on R, G, B... usually at about 80%
6. For R, G, and B do linear fit against the image with the lowest mean
7. Combine RGB
8. Image solve RGB combined if needed
9. SPCC... rarely fails
10. SCNR at 80%
11. Stretch, usually with GHS; followed by curves if needed for color and/or lightness. Sometimes use maskedstretch or EZsoftstretch or STF/HT
12. Luminance... same general process as R, G and B
13. Apply Local Histogram Equalization and or Unsharpmask to improve contrast of luminance. May use Advsharp script to sharpen.
13. Combine L and RGB after a slight blur of RGB... can remove a multitude of issues in RGB
14. Often use one of Bill Blanshan's star reduction processes
15. Photoshop editing if needed... usually not much

I would certainly appreciate feedback on this process.

that was really good, thanks George.

I simplified it a bit by not denoising each RGB channel, but ran noiseXterminator on the RGB image.
When integrating L it is key that the histogram of L has its peak exactly where the RGB is, otherwise colors go wild.

Here is my first attempt to grapple with RGB on a mono camera plus adding L and Ha channels: https://www.astrobin.com/7uajtj/E/
- as I keep taking images while I am learning to process, integration time is adding up nicely :-)

Arny

Arny… Thanks for the feedback and the points about incorporating luminance into the RGB. I'd be interested to hear how you prepare the luminance prior to integration into the RGB. I've tried using a synthetic luminance (when a real L is not available) and I'm amazed how much it can improve an RGB.

George

Hi mate - workflow looks fine for mono. 1 major point and 1 minor. minor - no need to linear fit, particularly if narrowband. if you're using a script or pixelmath for stretching it'll hit the same target background mean regardless of whether you linearfit or not.

major - if you're planning to use foraxx you have to stretch/delinearise the image before you combine, and if you apply blurx on a stretched image it'll sharpen only, not deconvolve. russell mentions this in his december 22 chat with adam block. you have to apply blurx then on the individual linear mono images before combining if you're using foraxx.

if you're using spcc remember it needs to be done at linear stage. my mono workflow for rgb is usually dbe (only if local normalisation didnt take care of it all), combine, spcc(make sure to set a background preview), blurx, stretch, starx, curves etc, noisex (if necessary), lum combine(hdrmt, lhe sometimes, contrast curves), more curves (mainly saturation, maybe more targeted contrast), re-screen stars.

i've found because I use script-based stretching that noisex at linear stage can give me some slightly misaligned histogram peaks if i stretch the denoised/destarred image. russel has said in interviews that he doesnt believe theres a material difference in outcome if you use noisex at linear or non-linear stages.

George Hatfield:
Arny... Thanks for the feedback and the points about incorporating luminance into the RGB. I'd be interested to hear how you prepare the luminance prior to integration into the RGB. I've tried using a synthetic luminance (when a real L is not available) and I'm amazed how much it can improve an RGB.

George

Hi George,

I haven‘t done synthetic luminance yet bit read, that L frames should make up 50% of total exposure time.

I prepare L by ABE or DBE, noiseXterminator, doing curves of masked core of the nebula and than aligning peak of histo with that of rgb before merging.

Arny

Hi mate - workflow looks fine for mono. 1 major point and 1 minor. minor - no need to linear fit, particularly if narrowband. if you're using a script or pixelmath for stretching it'll hit the same target background mean regardless of whether you linearfit or not.

major - if you're planning to use foraxx you have to stretch/delinearise the image before you combine, and if you apply blurx on a stretched image it'll sharpen only, not deconvolve. russell mentions this in his december 22 chat with adam block. you have to apply blurx then on the individual linear mono images before combining if you're using foraxx.

if you're using spcc remember it needs to be done at linear stage. my mono workflow for rgb is usually dbe (only if local normalisation didnt take care of it all), combine, spcc(make sure to set a background preview), blurx, stretch, starx, curves etc, noisex (if necessary), lum combine(hdrmt, lhe sometimes, contrast curves), more curves (mainly saturation, maybe more targeted contrast), re-screen stars.

i've found because I use script-based stretching that noisex at linear stage can give me some slightly misaligned histogram peaks if i stretch the denoised/destarred image. russel has said in interviews that he doesnt believe theres a material difference in outcome if you use noisex at linear or non-linear stages.

Thank you for the response. Understood on all your points. Last night I was working on each channel in linear form then stretched via GHS (I also removed the stars and stretched those as well). However when I ran a script for Foraxx I was not happy with the result or at least for my taste didn't result as I was hoping. I'm going to try your and some others on this thread workflow by combining the channels after DBE and going from there.

Hi George,

I haven‘t done synthetic luminance yet bit read, that L frames should make up 50% of total exposure time.

I prepare L by ABE or DBE, noiseXterminator, doing curves of masked core of the nebula and than aligning peak of histo with that of rgb before merging.

Arny... Thanks for the reply. Much appreciated.

**Update**

I started playing around with my data and after 2+ hours no final image. I ran into several problems, mainly, my Oiii and Sii are noisy. Both channels are about 5-6 hours.

I do think I made several errors a long the way. I used my Ha as luminance and I do not think I had matched its brightness to RGB. When I added the L layer my reddish/yellow Foraxx image turned bright pink (even when i just used it for sharpening on my second attempt extractong L from the stretched RGB image). I will work on aligning the histogram peaks and see if this gets me a better result.

I do think my data is not the best but this go was me learning how to shoot in mono and practice processing. Before I called it a night I combined the channels after dynamic crop and got the greenish hubble pallette. My plan for today is to DBE each channel and combine in linear then workup as an RGB image which would mimic my OSC workflow from before. I will work on luminance as the brightness was way too high and distorted my image vs sharpening. I'm trying to think what else could have gone wrong but I'll keep working on this.

Thanks to everyone for their advice and input

I've also switched to mono recently and had to develop a way to capture my target and process the data.
I use Siril and Starnet with the following steps for processing Ha and O-III data (should work for HSO as well):

Per filter: calibration, registration, stacking (using a script)
Common registration (alignment) of the resulting images from all filters
Per filter: BG extraction, optional deconvolution and star removal
Linear match of starless images (produce a balanced histogram)
RGB composition of the starless images (e.g. r,g,b=Ha,O-III,O-III)
RGB conversion of the starmask from one image (e.g. r,g,b=Ha,Ha,Ha)
Individual stretching of starless and starmask image
Denoise of starless RGB composition
Merging RGB composition with RGB starmask using pixel math (~(~StarMask * ~Starless))
Crop/saturation/export

An example that follows these steps can be found here: https://www.astrobin.com/7cwdm3/

I was fighting with pink/reddish RGB compositions for quite some time until I realized that balanced histograms are required.
Here is a previous image (my first attempt to process mono data) that shows what happens if the Ha image (becoming the r channel) has a higher level: https://www.astrobin.com/ypex30/

A more advanced variation of this process would be to capture a bit of R/G/B as well to get proper star colors.
Other targets will benefit from different processes (e.g. I would go with L/R/G/B for M42).

There are more complex processes (e.g. adding Ha/Hb to an (L)RGB image), but they need special pixel math treatment to avoid clipping or unbalanced colors (simply adding r=R+Ha and b=B+0.2*Ha will result an image with dominant reds). Also, I would only merge additional data into an image (or a process) that is already balanced.

Color calibration is an essential step in astrophotography processing, especially for deep-sky objects like galaxies and nebulae. It helps to remove the color cast that might be introduced by your camera's sensor. You typically perform color calibration after stacking your images and before any stretching or other processing. SPCC is another method of achieving similar results.Your workflow seems to be mostly focused on pre-stretch processing steps. You should consider inserting the color calibration step after Dynamic Crop and before DBE. This way, you'll correct color issues before further processing.

You mentioned that BlurX works better on color images. While that's generally true, you can still apply it after stretching your image. In fact, many astrophotographers apply noise reduction, including blurring techniques, after stretching to avoid losing faint details during the stretching process. So, it's okay to apply BlurX after stretching as long as it helps improve the image quality.

I have 2 general work flows depending on target and if I want to use Foraxx or not.

Standard palette (RGB/SHO) workflow (Linear);

Stack, Siril, APP or WBPP all work. I prefer APP
Pass each channel through GraXpert
Bring channels into PI
(Star Align and Normalize to brightest channel if stacked in Siril)
Channel Combine
Crop edges - Keep a copy of this stage for a star align if you use RGB stars or HOO stars in an SHO image. Also handy for a Lum or Ha continuum layer.
Plate Solve
SPCC (use NB mode for SHO)
BlurXterminator
StarXterminator
GHS or
Soft stretch (under stretch it) then
SCNR and/or Invert/SCNR x0.8 for magenta background
Curves for Saturation
Curves and/or GHS (great place to do some colour adjustment) for fine touch up stretch
NoiseXterminator
Out to GIMP/Affinity for Star recombine, cosmetic tweaks, crop, compress etc.

Foraxx or non-linear process

Stack, Siril, APP or WBPP all work. I prefer APP
Pass each channel through GraXpert
Bring channels into PI
(Star Align and Normalize to brightest channel if stacked in Siril)
Crop edges
BlurXterminator
StarXterminator
GHS or
Soft stretch (under stretch it) then GHS
Channel Combine or Foraxx Script
SCNR and/or Invert/SCNR x0.8 for magenta background
Curves for saturation/Fine stretch etc.
NoiseXterminator
Out to GIMP/Affinity for Star recombine, cosmetic tweaks, crop, compress etc.

For LRGB I will use the linear workflow for the RGB and non-linear for the Lum channel, same if I want to use the Ha as a Lum layer. I find that doing the NR very late in the process can prevent some blotchiness from creeping in during the stretch and over all give me much finer control over the details. For NB stars, I often find that I can get decent results by performing and SCNR and inverted SCNR. Not nearly as nice as well processed RGB stars but usable.

So many ways to arrive at the same place. I have no idea what is best and I'm still trying lots of different experiments to figure out the tricks for each image type.