Sharp Module Use (v1.6)

Notes from users, documentation addendums.

Sharp Module Use (v1.6)

Postby Guy » Wed Apr 01, 2020 9:04 am

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 version 1.6 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 easily access similar notes on the other StarTools modules see StarTools Main Window Use.

Sharp Module (V1.6)

  • To provide a range of tools which give the ability to sharpen the small to medium detail in the image.
For a general overview see Sharp: Multi-Scale Noise-Aware Structural Detail Enhancement
The Sharp module uses wavelet sharpening techniques to enhance the detail in the image.
The module is scale aware - it splits the image into multiple layers of different scale elements - allowing the enhancing of detail of particular sizes only.
The module is also noise-aware - allowing you to control its effect on areas depending on their SNR.
Only detail that has a high enough SNR is emphasised. Noise grain propagation is minimised.

Useful Sources
Sharp: Multi-Scale Noise-Aware Structural Detail Enhancement

When to use:
  • After the final global stretch (AutoDev or Develop), Contrast and HDR - if used.
  • Before the Decon Module.
  • Use only once.
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}
Key: {...} optional modules

This is a way of using the module which should give good results in most cases:
  1. Select the Structure Size you want to sharpen - and press Next
  2. Create an inverse star mask. Mask-Auto-Stars-Invert.
  3. Adjust the 'Amount' control to bring out the structures as much as possible without increasing the noise at all scales.
  4. Within the selected Structure Size are 5 different layers of different scale sizes. If necessary, reduce the individual Scale setting to reduce the effect at that scale.
  5. Select a Low SNR bias depending on the image type. Low (0) for DSO, High(around 85%) for Planetary.
  6. Toggle top "Pre Tweak/Post Tweak" button to see the effect of last adjustment if needed.
  7. Press 'Keep' when done.
What result to look for:
  • Look for structures becoming clearer - less blurred.
  • Watch for stars bloating. If this happens, improve the star mask to counter this effect.
  • Watch for noise being enhanced. If this happens at all scales then back off the Amount setting. If it is enhanced at a particular scale then back off the relevant Scale setting instead.
After Use:
  • Use the Life module if needed.
Special Techniques:
Sharp module can also be used as a scale decomposition tool - being able to remove or attenuate features of a particular size. This allows different versions of the image with different scales to be processed separately and blended together at a later stage.

Description of Controls:

Screen 1:

Structure Size:
Specifies the largest (Scale 5) size the wavelet sharpening should cover:
  • Large - elements of approximately 100-120 pixels downwards.
  • Medium - elements of approximately 40-50 pixels downwards.
  • Small - elements of approximately 10-15 pixels downwards.
  • Default is 'Large'.
  • Set this to the size of the largest element you want to sharpen (e.g. spiral arms, dust lanes).
Screen 2:

These set the other setting values based on the type of target.
  • DSO - Optimises the Settings for DSOs
  • DSO Light - As DSO - but allows full local brightening to enhance contrast. Sets Dark/Light Enhance to 0%/100%
  • DSO Dark - As DSO - but allows full local darkening to enhance contrast. Sets Dark/Light Enhance to 100%/0%
  • Planetary - Optimises the Settings for Planetary and Moon.

For general instructions on using mask see Mask.
  • Select what parts of the image to sharpen. Should mask out all visible stars.
  • Create an inverse star mask. Mask-Auto-Stars-Invert.
Sets the amount of sharpening at each of 5 different scale ranges.
  • Default is 100%. Range is 0% to 100%.
  • Reduce scale from 100% at a particular scale if noise is being enhanced at that scale. If changed - this is usually only necessary at scales 1-2.
  • Scales do not have absolute limits to the range - its is more like a particular scale brings detail of a certain size into focus - and that other detail is out of focus to varying degrees depending on its size.
  • The following are broad guidelines:
    • The largest scale (Scale 5) is set by the Structure Size parameter.
    • The smallest size (Scale 1) is always around one pixel.
    • The intervening scale sizes increase exponentially.
Scale Descriptions:
  • Scale 1 - Controls fine feature sharpening - This scale covers the smallest features such as single pixel detail.
  • Scale 2 - Controls the amount of medium to small feature sharpening.
  • Scale 3 - Controls the amount of medium feature sharpening.
  • Scale 4 - Controls the amount of large to medium feature sharpening.
  • Scale 5 - Controls the amount of large feature sharpening.
Adjusts the strength of sharpening across all scales
  • Default is 100%. Range is 0% to 1,000%.
  • Increase to make structures clearer. Decrease if noise is being enhanced too much at every scale.
Dark/Light Enhance
Specifies how the module can improve contrast - by specifying whether it is by darkening the surrounding area, lightening the detail, or a mixture of both.
  • Increasing this value - towards 0%/100% - means the module will be able to lighten areas rather than darken them.
  • Decreasing this value - towards 100%/0% - means the module will be able to darken areas rather than lighten them.
  • Use low values (towards 0%/100%) to enhance faint nebulosity without darkening the surrounding area.
  • Default is 50%/50% (for DSO & Planetary presets, 0%/100% for DSO Light and 100%/0% for DSO Dark). Range is 0%/100% to 100%/0%.
High SNR Size Bias
Specifies how the module resolves any conflicts between scales in high signal-to-noise ratio (SNR) areas.
Foe example, it may be that sharpening in a larger scale causes contrast reduction in a smaller scale.
  • The larger the value, the more important is the small-scale detail compared to the larger scale.
  • Normally the emphasis would be on the small-scale detail.
  • Default is 85%. Range is 0% to 100%.
Low SNR Size Bias
Specifies how the module resolves any conflicts between scales in low signal-to-noise ratio (SNR) areas.
  • The larger the value, the more important is the small-scale detail compared to the larger scale.
  • For DSO's, in low SNR areas is is more likely that only large scale detail can be recovered - so this should be low (0%).
  • For Planetary images, the small-scale detail should be protected so use a high value (around 85%).
  • Default is 0% (0% for DSO presets, 85% for Planetary preset). Range is 0% to 100%.
Mask Fuzz:
If a mask is used, Mask Fuzz controls the blending of the transition between masked and non-masked parts of the image.
  • Active if a mask is used to selectively sharpen the image.
  • Using this control will ensure smooth transitions between sharpened and unsharpened parts of the image.
  • Default is 4 pixels. Range 0 to 10 pixels.
Background Notes:
The Sharp module uses Wavelet (aka Laplacian) Sharpening techniques to enhance the detail in the image. See How Unsharp Masking and Laplacian Sharpening Work

Scale and regions affected
  • The scale is based on the number of pixels - so if you use the same settings on an un-binned image and the same image 50% binned they will have different scales affected.
  • For each scale increment the size increases exponentially, based on Structure Size. For Small the next scale size is 2x the current one, For Medium it is 2.75x and for Large it is 3.5x.
  • If you want to see what size a particular scale affects - set that Scale to 100%, the Amount to 1000%, and all other scales to 0%. Do Before/After to see the scale affected.
Scale aware processing
If you wondered how StarTools detects the different scales - here's is generally how it is done 'behind the scenes':
  • Make two (or more - depending on how many scales you want) copies of the image.
  • Blur one with, for example, a Gaussian blur.
  • Subtract the blurred image from the non-blurred image. The blur will have killed most of the fine detail (it's a blur after all), so subtracting the blurred image from the non-blurred image logically leaves you with just the fine detail.
  • You can now manipulate that fine detail - for example normalise it, use brightness/contrast operations on it, etc.
  • Add it (the modified fine detail) back to the blurred (coarse detail) image and voilà; you have an image that has had just a particular scale manipulated.
You could even do all this in something like The GIMP or Photoshop with the caveat that some additional trickery is required to manipulate and visualise pixels with a negative value.
Now, nothing of course stops us from grabbing that fine detail we isolated, perform the same trick on that, so that you can start getting "bands" of different detail sizes.
The radius of the kernel you use for the Gaussian blur in this case defines the image scale peak response.
Posts: 104
Joined: Thu Feb 19, 2015 8:35 am

Return to User Notes

Who is online

Users browsing this forum: No registered users and 1 guest