1.6.388; new Decon module!

[admin]

Hi all,

1.6.388 has a new Decon module, which takes things one step further again in terms of quality.
Again, the improvements rely heavily on the signal evolution Tracking data; a new regularization algorithm that continuously backward and forward propagates is now available for those who have systems that can handle it.
New is also the ability to sample and shape a suitable star as a Point Spread Function (PSF), and the ability to choose different PSFs to model an atmosphere (or no atmosphere).
To top it off, the Decon module is even capable of taking into account how your star sample is affected by each deconvolution iteration, tracking, re-sampling and re-shaping it as the image gets progressively deconvolved.
The docs (which can now also be launched in your web browser directly from StarTools by using the help button) have been updated, but a new video may be in order to demonstrate all the new features.
Even with all this new power, the module is still very easy to use. And if you were totally happy with the Decon module as it was, you can completely ignore all the new goodies (except for some renaming) and just keep using it the way you're used to (e.g. with a Gaussian).

Give the new version a spin and let me know if you find any issues. Thank you!

[almcl]

Ivo
Thank you so much for this latest release. Been having a play and finding more detail in previously processed images. Great!

One observation, if I may?

In previous versions, where Decon came before Sharp, the very useful Autogenerate mask function could be used to create an inverse star mask for both Decon and Sharp. Now, however, Sharp comes first in the work flow and that handy mask generation feature is not available until afterwards. Lazy, I know, but could it be made available in the sharp module, as well/instead?

[AndyBooth]

Fantastic work Ivo and team.
I asked about psf use in the module a very long time ago,
But it was not possible then.
It really makes a difference.
Also the new noise reduction features work great as well

[admin]

Ivo
One observation, if I may? In previous versions, where Decon came before Sharp, the very useful Autogenerate mask function could be used to create an inverse star mask for both Decon and Sharp. Now, however, Sharp comes first in the work flow and that handy mask generation feature is not available until afterwards. Lazy, I know, but could it be made available in the sharp module, as well/instead?

You raise a good point!
The way masks are used internally, have changed a little for both modules.
For the Sharp module it is best to create a mask that contains all (currently visible) stars.
For the Decon module it is best to create a mask that contains only over-exposing parts (such as fat stars); you will notice it is no longer referred to as a star mask in the docs, but rather a "Defect and singularity mask". The de-ringing code has had an upgrade as well in the Decon module. This means that, for the Decon module, you may even get away with not having a mask at all (use of the "Bright Response" slider may help here as well).

I need to think about what the most convenient/useful solution is here...

[admin]

Fantastic work Ivo and team.
I asked about psf use in the module a very long time ago,
But it was not possible then.
It really makes a difference.
Also the new noise reduction features work great as well

Thank you! I've thought long and hard about how to design a module that can use (useful) custom PSFs while still being effective and and easy to use. Besides the deconvolution regularization and algorithm, generating a suitable PSF is an art in itself. The modulating of the custom("secondary") PSF by the primary really helps with quickly finding the sweetspot in terms of "customness" and synthetic/theory-based atmospheric (or lens) modeling.

[elpajare]

Although it is offtopic, this version is faster than the stable one.

Congratulations

[admin]

Although it is offtopic, this version is faster than the stable one.

Congratulations

Thank you! A great number of optimizations went into 1.6. Most efforts went into Improved CPU data cache usage, improved muli-threading and reducing disk access during Tracking. The usage of AVX2 instructions may also help performance.

[alacant]

Hi everyone

For the Sharp module it is best to create a mask that contains all (currently visible) stars.

For Contrast too?

Great work with the Decon; a few examples of 'Primary PSF' and 'Secondary PSF' and where to use which would be ideal.
Cheers,
Steve

[AndyBooth]

Great for lunar shots too!
C11 EdgeHD mounted on a Skywatcher AZ-EQ6, close ups with 2x Barlow, Altair Astro GPCAM2 AR0130, with red filter.
Firecapture, Astrostakkert! 3, Startools.

[admin]

Great for lunar shots too!
C11 EdgeHD mounted on a Skywatcher AZ-EQ6, close ups with 2x Barlow, Altair Astro GPCAM2 AR0130, with red filter.
Firecapture, Astrostakkert! 3, Startools.

Nice results! I've been getting very, very good results by modelling atmospheric turbulence using the synthetic Moffat PSF as proposed by Trujillo et al. It seems to produce the most consistent results (with least amount of ringing) across a wide range of test images - deep sky, lunar, solar and planetary.
I've been looking for some nice lunar, solar and planetary data for the documentation, so if you (or anyone else), would like to donate (with full credits of course), please let me know!

For the Sharp module it is best to create a mask that contains all (currently visible) stars.

For Contrast too?
Great work with the Decon; a few examples of 'Primary PSF' and 'Secondary PSF' and where to use which would be ideal.

The Contrast module does not use masks (it operates on medium to large scale detail) and is not easily influenced by stars.
For some pointers regarding primary and secondary point spread function, how they work and what to look out for, please consult the docs.