Drizzle was originally developed for the NASA’s Hubble Deep Field observations. It provides a flexible and efficient way of combining dithered data, preserves photometric and astrometric accuracy, optimal signal-to-noise ratio and approach the best resolution that can be reached with linear reconstruction. (from “A method for the Linear Reconstruction of Undersampled Images” by A. S. Fruchter and R. N. Hook)
I’m so excited about this method… The results are fantastic compared to standard integration. Sharper and more details. Especially small objects taken at short focal lengths benefit from this method.
Until now I’ve drizzled the following images (see full size images!!!):
The colors looks different on M81, because I processed/changed colors in the past to reach the well known colors as much as possible. I recently decided to stop that. In the future I’ll only play around with the saturation.
Comparison of different integration methods
All 4 versions are processed exactly the same way (identical values/settings):
ChannelExtraction -> LinearFit -> ChannelCombination -> BackgroundNeutralization -> ColorCalibration -> SCNR (green) -> HistogramTransformation -> MorphologicalTransformation (StarMask) -> CurvesTransformation (RangeMask) -> UnsharpMask (RangeMask)
The drizzled version (based on Median/LinearFit integration) seems to have fewest background noise, M57 has more intense colors, higher contrast and a clear green-blue center. Overall an increased SNR (Signal-Noise-Ratio). Great!
Drizzle Processing in 3 Steps
- To simplify the process I use BatchPreprocessing to generate calibrated, debayered and registered images. These steps also can be done manually.
Add registered images and Drizzle files (created during step 1) , select combination method (Median) and check “Generate drizzle data” to prepare the drizzle files for integration (step 3 in this tutorial).
Warning: This step modifies the drizzle files generated in step 1! If you want keep them for later re-processing, make a backup copy now.
The result will be an image at double width and height (4 x larger image), ready for processing.
- Drizzle 2 x = 4 x larger image
- Drizzle 3 x = 9 x larger image
- Drizzle 4 x = 16 x larger image (…only with a super-computer 😉 )
Usually I process the full size 2x drizzled image like any other astro image (finally downsize in Lightroom to 50% to fit the “original size”). Better is to resample the image back to the original resolution in PixInsight (Process – Geometry – Resample)!
M57 standard integration: 5496 x 3670 px
M57 drizzle 2x integration: 10992 x 7340 px
- < 12 images: Sigma Clipping (or Percentile Clipping)
- 13-20 images: Winsorized Sigma
- > 20 images: Linear Fit Clipping