Russell Croman Astrophotography  

 

 

StarXTerminator

Home | Free Trial | Purchase | Support



Hover over the photograph to see the original


StarXTerminator is a Photoshop filter plug-in to remove stars from astronomical photographs. This allows separate processing of stars and background objects, or simply leaving out the stars altogether.

StarXTerminator uses an advanced convolutional neural network with a unique architecture suited to this task. This network has been extensively trained on photographs from both refractors and reflectors, so small stars, big stars, huge stars, and even diffraction spikes are recognized and removed by the network, with minimal impact to non-stellar features. StarXTerminator can even remove stars with dark halos around them from deconvolution or sharpening. An accurate feathering algorithm eliminates the tiling artifacts produced by other solutions. StarXTerminator can process many tiles in parallel, and automatically takes advantage of hardware such as GPUs and neural engines to accelerate performance.

The end result is a very smooth elimination of stars, with minimal residual artifacts and excellent color matching.

Recommended Usage

  • StarXTerminator works only on 16-bit RGB photographs. Convert your photograph to this mode if needed. Greyscale photographs can be processed by converting to RGB, running the filter, then converting back to greyscale.
  • For fastest performance, be sure to calibrate the Batch Size parameter for your system using the Calibrate button. This will perform an automatic search to find the batch size that results in the fastest performance possible on your system. This normally only needs to be done once, but should be repeated if you upgrade your system with more memory or acceleration hardware such as a GPU. StarXTerminator runs particularly fast on systems with supported GPUs or neural network accelerators (e.g., Macs with Apple Silicon).
  • Use StarXTerminator as early in your processing flow as possible. Though it has been trained to remove sharpened stars and stars with dark halos, better results are possible on lightly processed photographs.
  • Before running the filter, duplicate the target layer. This allows easy evaluation of the results as well as access to the original photograph in case any non-stellar features were removed.
  • If you want to create a layer with just the stars so they can be processed separately and added back in later, do the following:
    • Duplicate the target layer twice
    • Process the top-most layer with StarXTerminator
    • Duplicate this layer, then make it invisible
    • Select the first starless layer and set its blending mode to "Subtract"
    • Merge this layer with the one below – this is now the layer with just the stars
    • Move this layer to the top and set its blending mode to "Linear Dodge (add)"
    • Make the remaining starless layer visible again
    • You should now have three layers: the original photograph, the starless photograph, and the stars themselves

System Requirements

Please see current known issues and planned improvements on the support page linked above.

  • MacOS
    • Version 10.15 (Catalina) or later
    • Photoshop CC
    • GPU/Neural Engine recommended for fast performance, but not required. Not all GPUs are supported. StarXTerminator will automatically use the CPU if the GPU is not supported.
  • Windows
    • Windows verson 8.1 or later
    • Photoshop CC
    • GPU or other neural network acceleration hardware recommended for fast performance, but not required. StarXTerminator will automatically use the CPU if the GPU is not supported.

Version History

Version Date Comments
1.0 12 Sep 2021 Initial release

Is this StarNet++?

In case you are wondering if StarXTerminator is just the popular StarNet++ repackaged as a Photoshop plug-in, it is most definitely not. The neural network architecture, the training data set, the training method, and the supporting algorithms for tiling and batch processing are completely different and unique. StarXTerminator outperforms StarNet++ in every respect.