Compose Module Use

Notes from users, documentation addendums.
Post Reply
Guy
Posts: 119
Joined: Thu Feb 19, 2015 8:35 am

Compose Module Use

Post by Guy »

Compose Module

Here are some notes relating to using this module. It is not the only way to use it and experimentation is encouraged.
They relate to StarTools version 1.5 and later.
Please let me know if anyone sees any errors or has any additional advice they think helpful.
I will update this post as needed.
To see a full alphabetical list of module topics click here.

Purpose:
  • To allow the loading of separate files representing Luminance and Red, Green and Blue colour channels to enable complex composites.
Description:
The Compose module is used to load and combine multiple files into a single image. Either standard luminance, Red, Green and Blue data or a combination of narrow-band data using different filters.
This has a number of useful functions:
  • It allows simple creations of complex composites such as LLRGB
  • It can also be used to extract the Red, Green or Blue channel from a colour image.
  • If a channel is missing it can interpolate from the other channels to create the missing channel.
  • The balancing (weighting) of the different channels is done automatically based on the total exposure time for each channel.
While Tracking is on the 'Compose' button is lit up to show processing is being done in a special 'Compose Mode'.

In the 'Compose Mode' luminance and colour datasets are processed separately but in parallel in the subsequent modules - Bin, Wipe etc. - until Tracking is turned off.

Useful Sources
The processing video StarTools complex LLRGB composite processing in 9 minutes real-time, with the Compose module shows the use of the Compose module. This shows v1.5 but is still relevant for this module.

The notes below relate to StarTools version 1.5, 1.6, and 1.7

When to use:
  • At the start of processing if you need to load files representing different channels - either traditional LRGB or other combinations such as Hubble Pallette (SII, Ha, OIII)
Example Workflow (v1.7):
{Compose}-AutoDev-{Lens}-Bin-Crop-Wipe-AutoDev (or FilmDev)-{Contrast/HDR/Sharp/Decon/Flux}-Color-{Shrink/Filter/Entropy/SuperStr}-Track/NR-{Layer/Heal/Repair/Synth/Stereo 3D}
Key: {...} optional modules

Method:
  1. Preparation:
    • All the image files must be aligned (registered). When stacking, use one image as a reference for the others.
    • If a luminance file has been loaded, any Red, Green or Blue file subsequently loaded must be either exactly the same size or exactly half the dimensions of the luminance file. This allows luminance channels to be 1x1 binned while R,G and B files can be 2x2 binned. In all other cases all loaded files must be the same dimensions.
  2. The Compose module can be set to interpolate any missing channels. Just set Channel Interpolation on and load what channels you have.
  3. If you have any Luminance data, load that first. If you don't you must specify a Synthetic Luminance option in 'Luminance, Colour' setting otherwise the 'Compose' mode will not be enabled.
  4. Load any Red, Green and Blue data you have.
  5. Set the 'Luminance, Colour' parameter to define how the channels should be combined and what output you want. Common choices are: 'L, RGB' for no combination or 'L + Synthetic L from RGB, RGB' for LLRGB.
  6. If there is a difference in exposure time between Red, Green and Blue data set the Ratio settings accordingly.
  7. If you have problems loading data because of the error: 'The dimension of the bitmap you're trying to load differ from the previously loaded file bitmap(s). ...' then check the dimensions of individual data files by loading them in StarTools individually. The dimensions are listed to the right of the file name at the top of the screen.
What result to look for:
  • Make sure all the channels are properly aligned. In rare cases it is possible for the registration to be out. See the Background Notes section below on how to align files during stacking.
After Use:
Continue with your preferred workflow.

Special Techniques:

Loading Narrow Band data using the Hubble Palette
  1. Load the Luminance data - if you have it - into the Luminance channel.
  2. Load the SII data into the Red channel.
  3. Load the Ha data into the Green channel.
  4. Load the OIII data into the Blue channel.
  5. If you want to create a weighted synthetic luminance channel set the 'Luminance, Color' to 'L + Synthetic L from RGB, RGB'
  6. Otherwise if you have Luminance data (and don't want to use LLRGB) set the 'Luminance, Color' to 'RGB, RGB'
  7. Set the Total exposure times for each channel.
  8. Press 'Keep' when done.
There is a useful video Painlessly re-creating the iconic Hubble "Pillars of Creation" in StarTools.
See also the description: LRGB & Hubble palette

Simulate RGB image with SHO data
If you want to get a close as you can to a full spectrum RGB image similar to that produced by an OSC camera then use the following method as described in this post: Simulate RGB image with SHO filters
  1. Launch the Compose module
  2. Create the Luminance channel from S,H & O:
    • Load the SII data into the Red channel.
    • Load the Ha data into the Green channel.
    • Load the OIII data into the Blue channel.
    • Set the 'Luminance, Color' to 'L + Synthetic L from RGB, Mono' to make a mono blend.
    • Press 'Keep' when done.
    • Save this result - this mono file will be used for the SHO synthetic luminance.
  3. Create a red channel from S+H:
    • Re-launch the Compose module
    • Load the SII data into the Red channel.
    • Load the Ha data into the Green channel.
    • Ensure 'Luminance, Color' is set to 'L + Synthetic L from RGB, Mono' to make a mono blend.
    • Press 'Keep' when done.
    • Save this result - this mono file will be used as the red S+H channel.
  4. Combine these files with O in G and B channels to get the complete image:
    • Re-launch the Compose module.
    • Load the SHO synthetic luminance file in the Luminance channel
    • Load the S+H red file in the Red channel.
    • Load the OIII data into the Green channel.
    • Load the OIII data into the Blue channel.
    • Set 'Luminance, Color' to 'L, RGB'.
    • Press 'Keep' when done.
  5. Process as normal.
Creating a Synthetic Luminance channel from Wide Band data
  1. Load the Wide band image data into the Red, Green and Blue channels.
  2. Set the 'Luminance, Color' to 'L + Synthetic L from RGB, RGB'
  3. Set the Total exposure time for each of the Red, Green and Blue channels.
  4. Press 'Keep' when done.
Creating a Synthetic Luminance channel from OSC/DSLR Wide Band data
Note:
i) As of StarTools v1.5.366 this is done by default to standard OSC/DSLR data that is not white-balanced - there is no need to use the Compose module to get the advantages of Compose mode.
ii) As of 1.7.420 just use the 'Luminance, Color' setting 'L + Synthetic L from R(2xG)B, RGB (Color from OSC/DSLR)'. Keep Green channel duration the same as the others.
Before 1.7.420:
In this case the data has been debayered - and you need to adjust the total exposure time to compensate for the effect of the Bayer matrix.
  1. Load the Wide band image data into the Red, Green and Blue channels.
  2. Set the 'Luminance, Color' to 'L + Synthetic L from RGB, RGB'
  3. Set the Total exposure time for each of the Red, Green and Blue channels - set the Green channel to double the others (assuming RGGB Bayer matrix).
  4. Press 'Keep' when done.
Re-centering blue back onto stars
To re-centre a smeared blue channel back to align with the red and green channels try the following as described in Fringe killer filter add blue back to central star
  1. Optionally Bin the image - then save it.
  2. In the Compose module click Blue and navigate to the file just saved. This will load only the Blue channel of the saved image. The other channels will be automatically set due to 'Channel Interpolation'. You end up with a black & white representation of the blue channel
  3. Process this image so that you reduce the bloating of the stars in the blue channel
    • Do AutoDev or Develop - this allows us to see what we are doing.
    • You could then use the Decon module - set to 'De-ring Mask Gaps, Show Result', set Iterations to 1, set Radius to 3.2 (as appropriate).
  4. 'Keep' the result.
  5. Use Restore selecting 'Linear, Wiped, Deconvolved' option - this reverts to an image which is Linear but which retains the deconvolution you did.
  6. Save the image.
  7. Use the Compose module to load the Red and Green channels (also 50% binned) and load the saved image into the Blue channel.
  8. You should now have an image with a better focused blue channel. Process as normal.
  9. When using the Color module, the Scientific mode may have problems with the star cores as they have artefacts in the blue channel due to the deconvolution. Either:
    • Reduce Bright Saturation right down and set Saturation Amount to 100% or lower. This will avoid recovering colour in the highlights, or
    • Use a star mask to mask out the stars - they then won't be involved in setting the colour balance, or
    • Use Artistic mode
Adding Ha to OSC data.
  1. Create new red channel:
    • Load Compose Module.
    • Channel Interpolation On.
    • Load full-colour data into red channel Only using the Red/S-II button
    • Save that dataset - this creates a mono dataset of the red channel.
  2. Create a blend of this with Ha:
    • Load Compose Module.
    • Load these two datasets - the red channel data using Red/S-II button, and the Ha data using Green/Ha button.
    • Channel Interpolation On.
    • Set 'Luminance, Color' to 'L + Synthetic L from RGB, Mono' to create a mono blend.
    • Save this blend - this is the new red channel.
  3. Combine this new red channel with the other color data:
    • Load Compose Module
    • Set 'Luminance, Color' to 'L + Synthetic L from RGB, RGB'
    • Load new red channel using the Red/S-II button.
    • Load full-color data set using the Luminance, Green and Blue buttons. You now have R+Ha, G and B.
  4. Process as normal.
Creating Bi-Colour images using Steve Cannistras method
This is a description of a method proposed by Steve Cannistra Modified Bicolor Technique for combining Ha and OIII images which can produce fairly 'natural' results. This is effectively Ha:(Ha+OIII)x1.5:OIII.
You may get better results if you preprocess the images separately and then combine them as described below.
  • Create Synthetic Green Image:
    • Load the Compose Module.
    • Load the Ha data into the Red/S-II channel.
    • Load the OIII data into the Green/Ha channel.
    • Press 'Keep' when done.
    • Save this result - this mono file will be used for the synthetic Green channel.
  • Combine together:
    • Load the Compose Module.
    • Load the Ha data into the Red/S-II channel.
    • Load the OIII data into the Blue/O-III channel.
    • Load the synthetic Green data into the Green channel.
    • Set the 'Luminance, Color' to 'L + Synthetic L from RGB, RGB' to make a mono blend.
    • Set the Green Total Exposure so that the Green channel has 1.5x the weight of the other channels. If the others have 60 mins exposure, reduce the Green Total Exposure to 60/1.5 = 45 mins.
    • Press 'Keep' when done.
Process as normal.

Description of Controls:

Luminance button
Allows you to navigate to and load the Luminance file.

Red Button
Allows you to navigate to and load the Red channel file.
  • If a luminance file has been loaded then any Red, Green or Blue file subsequently loaded can be either the same size or exactly a quarter of the resolution of the luminance file. This allows luminance channels to be 1x1 binned while R,G and B files can be 2x2 binned. In all other cases all loaded files must be the same dimensions.
  • If not loading visible (RGB) data - load the data relating to the longest wavelength into the Red channel. E.g for SHO load S-II as Red.
  • If loading a colour (RGB) image - extracts the Red channel.
Green Button
Allows you to navigate to and load the Green channel file
  • If not loading visible (RGB) data - load the data relating to the mid-length wavelength into the Green channel. E.g for SHO load Ha as Green.
  • If loading a colour (RGB) image - extracts the Green channel.
Blue Button
Allows you to navigate to and load the Blue channel file
  • If not loading visible (RGB) data - load the data relating to the shortest wavelength into the Blue channel. E.g for SHO load O-III as Blue.
  • If loading a colour (RGB) image - extracts the Blue channel.
Luminance, Color
Defines how to combine the data from each channel:
  • RGB, RGB (Legacy Software) - Uses the luminance from the R,G and B channels, and the colour from the R,G and B channels - legacy behaviour.
  • RGB, Mono - Uses the luminance from the R,G and B channels, does not output colour.
  • L, RGB - Uses the luminance from the L channel, and the colour from the R,G and B channels.
  • L + Synthetic L from RGB, RGB - Creates synthetic luminance data from the R,G and B channels and adds it to the luminance channel, and the colour from the R,G and B channels. Also known as LLRGB.
  • L + Synthetic L from RGB, Mono - Creates synthetic luminance data from the R,G and B channels and adds it to the luminance channel, does not output colour. This is useful for creating blends of different filtered data sets e.g. Ha+OIII. Save the file and then combine in Compose.
  • L + Synthetic L from R(2xG)B, RGB (Color from OSC/DSLR) - Creates synthetic luminance data from the R,2xG and B channels combined, and the colour from the R,G and B channels. LLRGB using OSC/DSLR (with bayer matrix) (v1.7.420+).
  • L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR) - Creates synthetic luminance data from the R,2xG and B channels combined, does not output colour. Supports OSC/DSLR (with bayer matrix) (v1.7.420+).
  • L + Synthetic L from RGB, R(GB)(GB) (BiColor) - Creates synthetic luminance data from the R,G and B channels combined, and the colour from the R and either G or B channels. Useful for creating Bi-colour images from 2 narrowband filtered datasets (v1.7.420+).
  • L + Synthetic L from R(2xG)B, R(GB)(GB) (BiColor from OSC/DSLR) - Creates synthetic luminance data from the R,2xG and B channels combined, and the colour from the R and G+B and G+B channels (Standard HOO palette). Palette can be changed in Color Module later. Useful for creating Bi-colour images from 2 duo/tri/quad-band filtered datasets. (v1.7.420+).
Channel Interpolation
The Compose module can be set to interpolate any missing channels.
  • Just set Channel Interpolation On and load what channels you have. This feature can be used to:
    • Generate a missing green channel in the case of an Ha/Hb composite
    • Generate a greyscale from a single Ha, Hb, OIII or SII frame which may later be turned into a false colour image using the Color module.
  • Default is On
Luminance File
Shows the path of any file loaded in the Luminance channel.
  • Default is None.
Red File
Shows the path of any file loaded in the Red channel.
  • Default is None.
Green File
Shows the path of any file loaded in the Green channel.
  • Default is None.
Blue File
Shows the path of any file loaded in the Blue channel.
  • Default is None.
Lum Total Exposure
Sets the total exposure time of the luminance data.
  • Default is Not Set. Range is 1 minute to 720 minutes.
Red Total Exposure
Sets the total exposure time of the red data.
  • Default is Not Set. Range is 1 minute to 720 minutes.
Green Total Exposure
Sets the total exposure time of the green data.
  • Default is Not Set. Range is 1 minute to 720 minutes.
Blue Total Exposure
Sets the total exposure time of the blue data.
  • Default is Not Set. Range is 1 minute to 720 minutes.
Background Notes:

Compose mode
In 'Compose Mode' the Wipe (and other) modules works on the luminance and colour datasets in parallel.
You can see the results of each by pressing the 'Color/Luminance' button in Wipe.
Remember to check both results.

As of StarTools v1.5.366 standard OSC/DSLR data that is not white balanced uses the Compose mode by default.

Processing luminance and colour data separately for better results
The Compose module has at its core the idea of processing luminance and colour data separately. The luminance and colour data are split in the Compose module and recombined in the Colour module.
This is done because in many cases in post-processing the process is best applied to the luminance and not the chrominance (colour):
  • By applying some of the processes to the colour data as well as the luminance it is possible that you end up enhancing the noise. See the PixInsight article Why Separate Luminance and Chrominance?.
  • In addition, our eyes are much more sensitive to the luminance than the colour of an image.
  • StarTools splits luminance and colour to reduce the negative impact of some of the processes on the colour of the image.
  • By separating the luminance and colour data it is possible to process the luminance to get the maximum detail without the trade-off of increasing the noise in the colour. In addition it is possible to apply more aggressive denoise in the colour to further reduce the colour noise without noticeable effect.
  • With Narrow-band imaging it is quite common to capture L,R,G and B data with more L data than R,G and B. The R,G and B is combined and processed as one frame. The luminance (L) data is processed separately and recombined with the RGB towards the end of Post-Processing.
  • It is possible to do the same with OSC and DSLR colour frames by separating out the L and RGB components.
  • See the Special Techniques sections both here and in the Layer Module Use.
    • Layer Module - 'Creating a synthetic luminance frame'
    • Layer Module - 'Splitting a colour image into luminance and RGB before Processing'
    • Layer Module - 'Combining luminance and RGB images using the Layer Module'
Combining data sets
There is an interesting musical analogy Ivo uses here
The summary is:
  • Broadband data - The Exposure sliders are for carefully controlling the mixing of the R G and B channels to create a synthetic luminance when the R G and B channels represent wide bandwidth data and the R, G and B combined cover the full visible spectrum. This retains the 'documentary' value of the image.
  • Narrowband data - We can use the Compose module to add Narrowband data in various ways in order to aim for certain effects. However, using Narrowband data in this way does not have any 'documentary' validity. This, however, may not be important - and the results can be stunning.
Combining OSC and Narrowband data
Interesting discussion about the different ways of combining Narrowband and RGB data here

Considerations:
Sometimes the Narrowband data coincides with obvious visual elements (e.g. in HII objects where the Ha signal is generated in the same region)
  • In this case using the data to add to luminance makes sense
  • or adding it to an appropriate R, G, or B channel depending on the nearest visual wavelength is (usually R channel for Ha)
And sometimes there are no prominent corresponding visual elements (as in the case with Ha in M33)
  • In this case processing the Narrowband element separately makes sense and will give full control of the impact on the image.
    To do this process the Narrowband separately (just Wipe and AutoDev may be enough) and combine it with the LRGB version in the Layer module.
Making multiple images the same size, resolution and alignment
Images loaded in LRGB must be the same pixel size and alignment.
In some cases you might be trying to combine images that are not aligned or not the same pixel size. For example if taken with different cameras, on different nights or with different filters.
There are a number of ways to do this depending on what the differences are:
  • Images are the same physical size and resolution but different alignments. Either:
    - Use Regim to align frames as described here - using a single frame or stacking multiple frames. The result will be a frame of the same alignment as the reference frame. or...
    - Use Deep Sky Stacker (DSS) to align frames as described here - using a single frame or stacking multiple frames. The result will be a frame of the same alignment as the reference frame.
  • Images are the same physical size and alignment but different resolutions. Either:
    - Use Regim to align frames using the lowest resolution frame as a reference. The result will be a frame of the same resolution as the reference frame. or...
    - Use the Bin Module in StarTools to reduce the resolution of the higher resolution frames to that of the lowest resolution frame. Save the image as a Tiff file to be loaded in the relevant channel later.
  • Images that are different physical sizes and possibly alignment:
    - Use Regim to align the frames using the lowest resolution frame as a reference. The result will be a frame of the same resolution as the reference frame.
    - Some cropping will be needed after combination to remove any borders.
Aligning LRGB images using Regim
To make sure the L, R, G, B images are all aligned the same. we need to select a single reference frame for all images.
  • In Regim images can have different dimensions. The output file dimensions will be that of the reference file.
  • This technique works when stacking multiple frames or with a single frame plus reference.
  1. Under the Preprocessing menu select the Preprocessing option.
  2. Set the standard Calibration, Register and Combination settings under the relevant tabs.
  3. Under the 'Files' tab select the luminosity reference light frame first.
  4. Select all other relevant lights, darks, flats etc.
  5. Select only the 'do calibration' and 'do register' checkboxes in the preprocessing dialog.
  6. Press OK to do the registration.
  7. All files registered to the reference frame will be .fit files and have the prefix “Reg_”. The reference file will not have one.
  8. Open the preprocessing dialog again and select all the "Reg_*.fit" files you want to combine and Add to images - exclude the reference frame if you want.
  9. Select only the 'do combination' checkbox.
  10. Press OK to do the combination.
  11. Save the image.
  12. Repeat for Luminosity, Red, Green and Blue data using the same reference frame.
  13. The result should be 4 files all aligned exactly the same.
Aligning LRGB images using Deep Sky Stacker (DSS)
To make sure the L, R, G, B images are all aligned the same. We need to select a single reference frame for all images.
  • In DSS all files to be stacked must have the same dimensions, number of colours, number of channels, only one master dark, offset and flat.
  • This technique works when stacking multiple frames or with a single frame plus reference.
  • Make sure you use the 'Standard Mode' setting in the 'Result' Tab of the Stacking Parameters in DSS.
  • If the reference sub-frame appears to be ignored if not included in the stack - try registering all the files again.
  1. Load the reference (luminosity) sub-frame into DSS.
  2. Right-click on it and select 'Use as reference frame'.
  3. Load relevant Lights, Darks, Flats etc. into DSS.
  4. Stack as normal and save the image.
  5. Repeat for Luminosity, Red, Green and Blue data using the same reference frame.
  6. The result should be 4 files all aligned exactly the same.
Colours and Filters
Eyes Response to Light
The generally accepted range of visible light is from 400nm-700nm although this varies from person to person and with different light levels. or more information see Eye spectral response.
  • Photopic: 50% 510nm-610nm, 5% 450nm-660nm
  • Scotopic: 50% 455nm-550nm, 5% 410nm-600nm
RGB Filters
The RGB filters commonly used have a spectral response as follows:
  • R: 600-700nm
  • G: 500-600nm
  • B: 425-500nm
Narrow-band imaging
Using different narrow-band filters to create three data sets and combining them using LRGB channels can have advantages:
  • Allows imaging through a lot of light pollution (other than LED broad spectrum light pollution).
  • Nebulosity appears more detailed.
  • Stars seem less bloated.
  • Colours are artificial - so can be chosen to highlight features - e.g. Hubble Palette better differentiates SII and Ha.
  • Common narrow band data collected is:
    • SII - Red - 671.9 nm and 673.0 nm
    • Ha - Red - 656.3 nm
    • NII - Red - 654.8 nm and 658.3 nm
    • OIII - Cyan - 495.9 nm and 500.7 nm
    • Hb - Blue - 486.1 nm
    • OII - Near U-V - 372.6 nm and 372.9 nm
  • Three colour narrow-band images:
    Common combinations for assigning the three bands to the R,G and B channels are:
    • SII : Ha : OIII (R:G:B) - Hubble palette.
    • Ha : OIII : SII (R:G:B) - Canadian-France-Hawaii Telescope palette.
  • Bi-colour narrow-band images:
    Bi-colour images sometimes have advantages over tri-colour images: Many possible bi-colour combinations of assigning the two sets of narrow-band data to the R,G and B channels - common ones are (R:G:B):
    • Ha : OIII : OIII
    • Ha : (Ha+OIII)x1.5 : OIII - Steve Cannistra's technique - see description in Special Techniques section above.
    • Ha : Hax0.4 + OIIIx0.6 : OIII
    • SIIx0.5+ Hax0.5 : Hax0.4 + OIIIx0.6 : OIII
  • Synthetic H-beta
    Wherever Ha appears there is often also Hb. You can capture the Hb with a filter but you can also easily create a synthetic Hb.
    See Synthetic RGB from [SII], Halpha and [OIII] Emission Line Data and Narrowband imaging with only H-alpha and OIII filters
    • The ratio of Ha:Hb emission is normally about 3:1 for emission nebulae and 6:1 for planetary nebulae.
    • The shorter wavelength Hb has a higher extinction so the perceived amount of Hb is lower.
    • Common to Combine OIII and Synthetic Hb into the blue channel as follows: OIII x 0.85 + Ha x 0.15 for an emission nebula with variation depending on the source.
  • For a list of targets see the CN list of advanced narrow band imaging targets.
Heno
Posts: 19
Joined: Wed Feb 17, 2016 8:07 pm

Re: Compose Module Use

Post by Heno »

Great write-up! :thumbsup:
One question.
Under preparation you wrote:
If a luminance file has been loaded, any Red, Green or Blue file subsequently loaded must be either exactly the same size or exactly half the size of the luminance file.
Further down under Red button:
If a luminance file has been loaded then any Red, Green or Blue file subsequently loaded can be either the same size or exactly a quarter of the resolution of the luminance file
Should't these read the same?

Clear skies
Helge
Guy
Posts: 119
Joined: Thu Feb 19, 2015 8:35 am

Re: Compose Module Use

Post by Guy »

If the Red, Green, or Blue file is half the size (i.e. half the no. of pixels in width and height) of the luminance then it will be a quarter of the resolution.
The Red, Green, or Blue channels are allowed to be 2x2 binned (i.e. a quarter of the resolution) with respect to the luminance channel.

I think the way I refer to the size of file is confusing - I will try to make it clearer.

Regards,
Guy
Post Reply