New version of the SubframeSelector tool (SubframeSelector module 1.4.8)

Juan Conejero

PixInsight Staff
Sep 2, 2004
Valencia, Spain
We have just released version 1.4.8 of the SubframeSelector module as an update for PixInsight core 1.8.8-8. The update is now available to all licensed users on Linux, macOS and Windows.

This new version solves a number of issues and provides several improvements to SubframeSelector, a critical component of our data reduction and preprocessing pipelines. This is the first of a series of updates planned for this important tool. This version introduces a few breaking changes that I'll describe succinctly below.

  • PEDESTAL keywords are now handled exactly the same way as in the ImageCalibration process, including a dedicated Pedestal section in the graphical interface. This means that existing PEDESTAL keywords will be subtracted automatically by default from all input frames.

  • In previous versions of the SubframeSelector process, the literal pedestal value was expressed in the range from 0 to the specified A/D resolution of the camera (for example, from 0 to 4095 for a 12-bit camera). Now pedestals are always expressed in the 16-bit unsigned integer range [0,65535] for coherence with the rest of the platform, irrespective of how they are specified, be it literally or acquired from metadata.

  • All subframes are now clipped to the following range of pixel sample values for median calculation:

    [m + 1.0/65535, M – 1.0/65535]

    relative to the normalized [0,1] range, where m is the minimum pixel sample value in the image and M is the maximum value, both expressed in [0,1]. This allows us to work with registered frames, where we must make sure that any black regions will be ignored (consider e.g. mosaic tiles), as well as with calibrated frames, including low-SNR frames with a significant amount of negative sample values after pedestal subtraction.

  • SubframeSelector acquires now noise estimates from existing NOISExx and NOISEFxx keywords, respectively for estimates of the standard deviation of the noise and fraction of noise pixels. Normally these keywords are generated by the ImageCalibration and Debayer processes. If the corresponding keywords are present, they will be used instead of having to calculate new estimates, as previous versions always did. This is a significant improvement, both in terms of execution speed and memory consumption.

Bug Fixes
  • Low-SNR frames where a significant amount of negative pixel values result from pedestal subtraction are now evaluated correctly. Original bug report:

  • Like
Reactions: acmalko