Page 1 of 1

Color Module Use

Posted: Mon Jan 30, 2017 10:40 am
by Guy
Here are some notes relating to using this module. It is not the only way to use the module and experimentation is encouraged.
They relate to StarTools versions 1.5 to 1.8
Please let me know if anyone sees any errors or has any additional advice they think helpful.
I will update this post as needed.
For an index of similar notes on the other StarTools modules see StarTools Main Window Use.

Color Module

Purpose:
  • To achieve a good colour balance that accurately describes the colour ratios that were recorded.
Important:
  • To get best results this module should be used on a colour calibrated monitor.
  • If you can't do that the 'Max RGB' button helps by showing the dominant channel (R,G or B) for each pixel. This can help you adjust the colour if you have an uncalibrated monitor.
Description:
For a general overview see Color: Advanced Color Correction and Manipulation.
Processing can modify the colour balance - in particular stretching the image to bring out detail can have a bad effect. This module helps correct those changes.

Useful Sources
There are very good general instructions for the use of the Color module in Color Usage.
Also here is another detailed description by Ivo: Color module demystified (v1.3.2+).
The Hangout discussion of StarTools with Ivo discusses the Color module between about 1h14m and 1h47m (v1.3.5).

When to use:
  • Towards the end of the workflow (after all the stretching) just before turning off Tracking and doing the final Denoise.
  • You can re-use this module if you use one of the 'LRGB Method Emulation' modes that uses CIELab Luminance Retention.
  • Re-use if you want to adjust the colour or saturation in different areas by using a mask.
Example Workflow (v1.5):
AutoDev-{Band/Lens}-Bin-Crop-Wipe-AutoDev(or Develop)-{Decon/Sharp/Contrast/HDR/Flux/Life}-Color-{Entropy/Filter}-Denoise-{Layer/Magic/Heal/Repair/Synth}

Example Workflow (v1.6):
AutoDev-{Band/Lens}-Bin-Crop-Wipe-AutoDev (or Develop)-{Contrast/HDR/Sharp/Decon/Flux/Life}-Color-{Entropy/Filter}-Denoise (or Denoise 2)-{Layer/Shrink/Heal/Repair/Synth/Stereo 3D}

Example Workflow (v1.7):
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}

Example Workflow (v1.8):
AutoDev-{Lens}-Bin-Crop-Wipe-AutoDev (or FilmDev)-{Contrast/HDR/Sharp/SVDecon}-Color-{Shrink/Filter/Entropy/SuperStr/NBAccent/}-Track/NR(Unified-Denoise)-{Flux/Repair/Heal/Layer/Synth}

Key: {...} optional modules

Method:
This is a way of using the module which should give good results in most cases. If you have used a narrow band or light pollution filter then you should look at the Special Techniques section:
  1. At startup, if there is a full mask set, the module auto-calibrates the white point in the image.
  2. Select the preferred 'Style' - choose from: Scientific, Artistic Detail Aware, or Artistic Not Detail Aware.
  3. Select the 'LRGB Method Emulation' you want to use - usually this is 'Straight CIELab Luminance Retention'.
  4. Select 'Matrix' option as appropriate (v1.6)
  5. See if you get what you expected (see below for 'standard' colouring).
  6. If there are problems:
    • Use MaxRGB to look for issues (like unexpected green-dominant areas) - adjust the Red, Green and Blue bias to get a better balance.
    • Use Cap Green control to eliminate any unwanted remaining green.
    • If there are colour problems around stars or other highlights - use Highlight Repair (v1.6).
  7. If you make a mistake, the 'Reset' button discards all the changes since you started using the module.
  8. 'Keep' the result when you are finished.
Sampling Methods:
Parts of the image can be sampled to establish a good colour balance. The following ways are available:
See also the topic Auto Color Balance and Color Balancing Techniques.
  • Sampling overrides the current settings.
  • Use Sampling when automatic colour balance is difficult due to, for example, noisy data.
  • Sampling doesn't work when a filter (e.g. a narrow band or light pollution filter) has been used as the stars are all missing part of their spectrum - so no good reference white objects, or broad range of spectral types, exist. See 'Special Techniques' for an alternative approach.
  • This colour balance technique is based on the assumption that the sampled object(s) is, on average, a good reference white so we can establish the correct relative balance of the R,G and B channels from this.
  • White light contains all the visible spectrum so we look for objects which contain the same broad range:
    • 'White' Galaxies and some Star fields contain a broad range of spectral class stars and so are considered good reference white.
    • Spectral class G2V stars, like our sun, are often considered suitable white references but some people argue that they don't have a unique role as reference source. See the discussion in the Background Notes.
  • Globular Clusters often don't have a good mix - they mostly have very old stars - mainly yellow with some orange and reds - so should be excluded from a star field calibration.
  • For a discussion of these colour balancing techniques see the following references:
    Getting the colors right in astrophotos.
    PixInsight color calibration methodology.
    See also this Starizona article on True Color Imaging.
Single sample
See also Setting white reference by clicking pixel.
  • Click on an area in the image that should be white.
Sample from Mask
See also Setting white reference by mask sampling.
  • Define the elements that make up your sample in the mask (e.g. a galaxy, G2V star, or star field with good mix of star temperatures).
  • Click the 'Sample' button - this uses the mask to define the sample from which the white balance is determined and the Red, Green and Blue bias settings are established.
Star Field Calibration - assumes the star field has a good mix of star temperatures
See also Starfield Colour Calibration.
  • Mask - Auto - 'Fat Stars' preset - Do - (optional grow 1 pixel) - Keep.
  • Click Sample (uses the star samples and sets RGB bias control settings based on that).
  • Mask - Clear - Invert - Keep.
  • The module remembers the RGB settings but now applies them to the whole image (based on the new mask).
What result to look for:
The colour distribution you are looking for depends to a large degree on your preferences.
(See also the topic M31 in Moonlight).
If you look at these Thumbnails of Images of M8 you can see the range of colours of the processed M8 images on the internet. Many of them have a red bias and this can be a side-effect of the non-linear stretching causing a colour skew that has not been compensated for.
If using the 'Scientific (Color Constancy)' approach you should look for the following:
  • Good distribution of star colours - Foreground stars should show a good distribution of colour temperatures from red through orange, yellow and white to blue.
  • If a light pollution filter is used the star colours may just be orange and blue with little in between - yellow is often missing.
  • Check for green - This should be rare unless there is an OIII emission region (e.g. M42 core or Tarantula Nebula).
  • The H-alpha should look red, H-beta should look cyan.
  • HII areas (H-alpha + H-beta) should look purplish/pink.
  • Galaxy cores tend to look yellow (older stars) and their outer rims tend to look bluer (younger stars & star formation).
  • Dust tends to let through lower wavelength light (if any) - mainly browns and reds.
If using the 'Artistic, Detail Aware' or 'Artistic, Not Detail Aware' styles you should look for the following:
  • Bright areas will be paler, less colourful than above.
If using narrow-band techniques - such as the Hubble pallette or similar:
  • The automatic colour balancing that occurs on loading the module will probably give odd results because there is so much of the spectrum missing.
  • The 'Artistic, Not Detail Aware' style may work best with narrow band data - experiment.
  • Set the 3 Bias Sliders back to 1.00 and manually adjust until you get the result you want. This is false colour so there is no 'right' result.
  • Ensure the relative strength of each channel highlights the detail you want.
Light Pollution filter - tends to have a dip where the yellows should be - so galaxy cores lose their yellow.

Ways of getting better results:
  • Try to correct any colour problems (such as those caused by light pollution and gradients) using the Wipe module first.
  • Help the Wipe module by using Flats.
  • Temporarily increase the Saturation Amount control (to say 300%) while working with colour to help when guaging colour balance.
  • If using a light pollution filter visually getting the right balance will be very difficult as there are parts of the spectrum missing. It will often show a lack of yellow and some green when properly colour balanced - There is a way around this - see the Special Techniques section below.
After Use:
  • Normally at this point you are ready to stop Tracking and to use the Denoise module.
Special Techniques:

Colour Balancing data filtered by a Light Pollution Filter
This approach is summarised in the article Colour balancing of data that was filtered by a light pollution filter.
  1. Colour balancing by sampling of filtered data will not give meaningful results.
  2. Shoot luminance data with the light pollution filter in place.
  3. Shoot colour data without the filter in place.
  4. Process both images separately.
  5. Combine in Layer module as described in the topic Mel 15 (in this example it combines luminance and colour for Ha and RGB but the techniques are the same).
Colour Balancing data collected using narrow band filters (e.g. Hubble Palette)
For cases where you have used the LRGB module to load the data collected using SII, Ha and OIII narrowband filters to the R,G and B channels respectively.
  1. This is not intended to be true colour so colour balancing becomes a matter of taste.
  2. Optionally create a weighted synthetic luminance frame - As described in the article: Ha,R,G,B -> Synthetic Luminance.
  3. Adjust the relative proportions of SII, Ha and OIII using the R, G and B bias sliders until you get a balance that you like.
  4. 'Keep' the result.
Adjusting Colour of Stars and other features separately
Sometimes you want to adjust the colour of the stars separately from the rest of the image:
  1. Create a Star Mask.
  2. In Color module click 'Sample' - this sets the colour balance assuming the average star colour is white.
  3. Invert the mask - this will apply the white balance to the other (non-star) features.
  4. Adjust the saturation of the other features.
  5. Make any other changes to the colour balance you want.
  6. Invert the mask - so it selects the stars again.
  7. Adjust the saturation of the stars.
  8. Make any other changes to the colour balance you want.
  9. 'Keep' the result.
Creating Solar images with false colour
This is a simple way to put (false) colour into your monochrome solar images as described in the forum post here:
  1. Process your monochrome solar image as normal
  2. When you get to the Color module:
  3. Set the Matrix parameter to "False Color: Solar" - this will create a yellow false colour image as a starting point.
  4. Adjust the bias sliders to change the hue. Reducing the green bias slider makes the yellow less prevalent and brings out the reds.
  5. Make any other changes to the colour you want.
  6. 'Keep' the result.
Description of Controls:

Top Buttons
  • Cancel - Exits the module - discarding any changes.
  • Keep - Exits the module - saving any changes.
  • Mask - For general instructions on using mask see Mask:
    • The mask can be used to select areas to sample as a colour reference - see Sampling Methods section above. Click 'Sample' when done.
    • The mask can also be used to selectively adjust the colour of areas in the image. To do this you need to clear the mask before starting the module and set it when prompted at the start.
  • Sample - Uses the current mask setting as a sample set from which the white balance is determined and the Red, Green and Blue bias settings are established.
  • Reset - Sets the controls to neutral settings. Saturation to 100%, Bright and Dark Saturation to 1.0, R,G and B Bias to 1.0, All other settings to defaults.
  • Constancy - Optimised preset settings to support the 'Scientific (Color Constancy)' style.
  • Legacy - Optimised preset settings to support the 'Artistic, Not Detail Aware' style.
  • Hubble - Optimised preset settings for narrow band data sets using the Hubble palette (v1.5).
  • SHO(HST) - Optimised preset settings for narrow band data sets using the Hubble palette - maps SHO:RGB (v1.6+).
  • SHO:OHS - Optimised preset settings for narrow band data sets blending channels to broadly map OHS:RGB(v1.6+).
  • Bi-Color (Duoband) - Optimised preset settings for narrow band HO data sets - maps HOO:RGB(v1.6+).
  • Max RGB - For each pixel, shows which channel - R, G or B - is dominant. See Max RGB Mode.
    • If your image is too red, pixels that are supposed to be 'neutral' (such as the background) will show mostly red. If your image is too green they will show mostly green. If, however, your image is well calibrated, these neutral pixels will alter between red, green and blue.
  • Before/After - Toggles the display of the image between the current and initial view.
Histogram:
The histogram shows the distribution of pixel intensity for the separate R,G and B channels.
  • Displays the pixels intensity distribution split into RGB channels.
  • Only pixels set in the current mask are counted.
Style:
See also the article Tweaking your colors.
Options: -
  • Scientific (Color Constancy) - keeps the colour regardless of brightness by separating luminance and colour processing.
  • Artistic, Detail Aware - Emulates much other software where bright areas can look washed out. Tries to compensate for local brightness manipulations during processing (e.g when using HDR), In these areas it will compensate for these changes and show more colour.
  • Artistic, Not Detail Aware - As above but does not try to compensate for local brightness manipulations during processing.
  • The Style control is only available when Tracking is engaged.
Saturation Amount:
Specifies the amount of colour saturation relative to the original image.
  • Default 200%, Range 0-1,000%.
  • Reducing to 0% turns the image monochrome.
Bright Saturation:
Specifies the colour saturation in the lighter areas.
  • Default is Full (10), Range 1.00-10.00 (Full).
  • Reduce this value where there are colour artefacts noticeable in the highlights. For example where there are colour fringes around bright star cores with one side blue and the opposite side red.
Dark Saturation:
Specifies the colour saturation in the darker areas.
  • Default 2.00, Range 1-10.
  • Reduce this value if there is a lot of colour noise in the dark background.
Cap Green:
  • Green is produced by OIII emission regions (e.g. M42 core or Tarantula Nebula) which are rare.
  • Stretching colour data with luminance causes a skew in the colour balance.
  • Very few objects in space are predominatly green when imaged in RGB. So if we find a green pixel, and we are sure the colour balance is right, we can assume any pixels that are green are made that way by noise so we convert them to something more natural like yellow or brown.
  • Set to 100% to remove green completely.
  • Default is 0%. Range is 0% to 100%.
  • Use when you are sure the colour balance is right.
  • Use as a last resort.
  • Do as a final adjustment.
LRGB Method Emulation:
See also the article LRGB Method Emulation.
With Tracking on, StarTools has from the start separated L and RGB (if the data was linear when imported).
Now is the time to combine them - there are a number of different approaches to combining them to choose from here:
  • Straight CIELab Luminance Retention - adjusts all colours in a psychovisually optimal way in CIELab space, introducing colour without affecting apparent brightness.
  • RGB Ratio, CIELab Luminance Retention - Applies the RGB Ratio technique, with luminance retention in CIELab color space being applied afterwards.
  • 50/50 Layering, CIELab Luminance Retention - Applies the 50/50 Layering technique, with luminance retention in CIELab color space being applied afterwards.
  • RGB Ratio - This uses RGB ratios multiplied by luminance in order to better preserve star colour saturation.
  • 50/50 Layering - Here the luminance is layered on top of the colour information with a 50% opacity.
  • The default is 'Straight CIELab Luminance Retention'.
Bias Slider Mode:
Sets whether the Bias sliders increase or reduce the channel influence.
See also the article Setting a colour balance.
  • Default is 'Sliders Reduce Color Bias'.
Bias Sliders:
These sliders set the colour balance. They can be adjusted manually or by one of the colour sampling techniques described above.
  • The value is a simple multiplier (for Bias Increase) or divisor (for Bias Reduce), in the linear domain, of the initial R, G or B value.
  • It is applied before any other processing (e.g. the remapping using the Matrix parameter)
Red Bias Reduce (or Red Bias Increase):
  • Controls the Red channel or, for example, SII if using SHO narrowband.
  • Default 1.00 (no reduction/increase), Range 1.00-20.00.
Green Bias Reduce (or Green Bias Increase):
  • Controls the Green channel or, for example, Ha if using SHO narrowband.
  • Default 1.00 (no reduction/increase), Range 1.00-20.00.
Blue Bias Reduce (or Blue Bias Increase):
  • Controls the Blue channel or, for example, OIII if using SHO narrowband.
  • Default 1.00 (no reduction/increase), Range 1.00-20.00.
Mask Fuzz:
  • If a mask is used, Mask Fuzz controls the blending of the transition between masked and non-masked parts of the image.
  • Default is 1.0 pixels. Range is 1.0 to 41.0 pixels.
Highlight Repair (1.6+)
Specifies the size of the area around highlights (e.g. stars) where inaccurate color data may exist and will be repaired.
  • Color inaccuracy may be due to channel alignment or debayering issues.
  • Increasing this value increases the area around the highlight which will be repaired.
  • Default Off (0 pixels), Range 0-10 pixels.
Matrix (1.6+)
Contains either:
  • a list of DSLR cameras - if the dataset is imported as 'Linear, from OSC/DSLR and not white-balanced'.
    • This allows selection of the right color correction matrix for your camera.
    • Options include cameras from Canon, Nikon, Olympus, Pentax, Samsung & Sony
  • a list of common narrowband channel blends to choose from - if the dataset is imported as 'Linear'.
    • This allows you to remap the channels blend you originally set in the Compose module.
    • Affects the RGB channels but not the luminance channel.
    • Listed as <initial Compose mapping> <new mapping> with proportions as percent.
    • Does not affect the mapping of the Bias sliders (e.g. if you used narrowband SHO mapping in Compose then Red Bias Reduce/Increase always controls SII)
    • Options:
      • Remap SHO to different blends in each channel e.g: SHO 40SII+60Ha,70Ha+30OIII,100OIII
        assumes Composed with R=SII, G=Ha, B=OIII - to be mapped to R=40%SII+60%Ha, G=70%Ha+30%OIII, B=100%OIII
      • SHO Composed as OHS e.g: SHO:OHS 100OIII,70Ha+30OIII,40SII+60Ha;
      • SHO Composed as HOS e.g: SHO:HOS 70Ha+30OIII,100OIII,40SII+60Ha;
      • Duoband e.g: O(H+O)H Duoband 50G+50B,50R+25G+25B,100R
      • Interpolate Green channel e.g: Interpolate G 100R,(25R+75B), 100B
      • Swap channels round e.g: RGB:RBG
      • Remove a channel e.g: RGB:RG0
      • Colouring solar (monochrome) images - use: False Color: Solar
  • Default is 'Identity (Off)' - no correction or remapping.
Background Notes:

Star Colours
Stars radiate similar to a black body - with the colour made from a continuous spectrum - but with absorption lines.
The peak output wavelength depends on temperature - appearing as red, orange, yellow, white & light blue
They are not the colours of the spectrum - there is no green, indigo (dark blue) or violet
ClassTemp KApparent ColourNotes
M<3,700Orange-RedOld stars
K<5,200Yellow-Orange
G<6,000Yellow-white
F<7,500White
A<10,000White-Blue
B<30,000Blue-white
O>30,000Light BlueYoung stars and star-forming regions

Gas Colours
Excited gases produce emission spectra - Narrow band
IDElementIonisationColourWavlength (nm)
HIIHydrogenSingleRed-Pink656.3 nm
Ha+HbHydrogenSinglePurple (Hb Blue)656.3 486.1 nm
NIINitrogenSingleRed654.8,658.3 nm
SIISulphurSingleRed671.9,673.0 nm
OIIIOxygenDoubleCyan495.9,500.7 nm
OIIOxygenSingleNear U-V372.6,372.9 nm

Colours in Galaxies
  • Galaxies often show a predominance of Red & Yellow stars in the core - which is where star formation started first. This is where the blue stars, being hotter, have burned their hydrogen quicker and exploded.
  • The outer regions are often bluer - here star formation started more recently - and may still be going on - and the blue hotter stars still exist.
'Real' colours in Astronomical Images
Checking colours are ‘real’:
  • Full range of star colours - Older (red, orange, yellow) and younger (white and blue)
  • Gas clouds - HII regions Red (Ha) or Purple (Ha & Hb), SII and NII Red, OIII cyan
There is an interesting discussion of the issue of 'real' colours in astrophotography in the article by Jerry Lodigruss Color in Astronomical Images.
Also there is another related discussion on Starizona True Color Imaging.
This discusses ways of getting a proper colour balance including using a white (G2V) star.

G2V vs Other sampling techniques
The article Getting the colors right in your astrophotos discusses other approaches as well as discussing the idea that a sun-centric view of color balance (Anthropocentrism) does not make sense and so methods other than using G2V stars to color balance should be used.
This PixInsight forum post 'About our color calibration methodology' describes the PixInsight approach to colour balancing - which they call 'spectrum-agnostic' or 'documentary' calibration methods.
They 'try to apply a neutral criterion that pursues a very different goal: to represent a deep sky scene in an unbiased way regarding color, where no particular spectral type or color is being favored over others'.
They advocate using light sources which include a good range of stellar populations and spectral types. Examples they suggest are:
  • A nearby galaxy with negligible red shift.
  • A sampling of a large number of stars - by averaging a sufficiently representative sample of stars there is no bias towards one colour.
Colour Balancing and Light Pollution Filters
Traditional light pollution filters filter out the band of wavelengths that some artificial lights (e.g. low pressure sodium lights) produce.
This means there is part of the spectrum missing (in the orange-yellow region) which means it is not possible to colour balance using this data - see the Special Techniques section for a way of dealing with this.
With the rise of more broadband lighting (LED, HPS and Metal Hydride) this approach is also becoming less effective.
Here is an Interesting approach to using narrow band filters to overcome light pollution Using Photometric Filters to Overcome Light Pollution.
Instead of a single LP filter, it uses 3 narrowband filters Ha (656nm x 20nm), sYel (550nm x 19nm) and sV (410nm x 16nm) assigned to R, G and B. The 3 channels are colour balanced by calibrating against a white source.

Factors affecting colour balance:
A large number of factors can affect the white balance (interstellar, terrestrial and equipment) and not all of these have a uniform effect across the image. This needs to be taken into account when choosing reference points.
  • Red Shift (varies across image)
  • Emission gases - causing enhancement in specific wavelength (varies across image)
  • Interstellar extinction - reddening due to scattering off interstellar dust or other matter. (varies across image)
    - Absorbs and scatters blue more than red
  • Atmospheric Extinction - scattering by atmospheric dust (varies with altitude of target)
    - Scatters blue more than red - reasonably predictable - lowest at zenith, increasing with zenith angle
  • Transparency - atmospheric moisture content, pollution, ash etc. - can cause colour shifts (varies across image)
  • Telescope Optical path (uniform across image) - including:
    • RGB Filter transmission profiles
    • Light Pollution Filter
    • Sensor sensitivity profile
  • Intensity variation over time - affecting images made with RGB filters - since R, G and B are captured at different times
  • Preprocessing (uniform across image)
Colours in Narrow-Band Imaging
A tri-colour image - such as a properly processed SHO (Hubble pallette) narrow-band image - will show the following colours.
  • Green shades indicate a relative dominance of H-alpha. (Ha mapped to Green channel)
  • Orange indicates a relative dominance of H-alpha and S-II but a relative paucity of O-III. (Mix of Red and Green channels showing, Blue has little impact)
  • Blue indicates a relative dominance of O-III and a relative lack of H-alpha and S-II. (O-III mapped to Blue channel)
  • Teal indicates a relative dominance of H-alpha and O-III, but a relative lack of S-II (Mix of Red and Blue channels showing, Green has little impact)
A bi-colour image will similarly have characteristics depending on the how the two data sets are mapped to RGB.
For an HOO mapping:
  • Deep Red indicates a relative dominance of H-alpha with little O-III. (Ha in the Red channel - other channels have little impact)
  • Deep Cyan indicates a relative dominance of O-III with little Ha. (O-III in the Green and Blue channels - Red channel has little impact)
  • White indicates a balance between Ha and O-III. (Data strengths balanced to make Red, Green and Blue channels equal)
  • So a subject where there are large areas where Ha and O-III are balanced will show as pale and washed out. Choosing the right mapping for the subject - or the right subject for the mapping - is critical.
  • Getting the colours 'right' can be tricky.
  • See this post: Style and LRGB Emulation in Bicolor