Author Topic: New Lanczos Pixel Interpolation Algorithm (ImageRegistration & Geometry modules)  (Read 9010 times)

Offline Juan Conejero

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 7036
    • View Profile
    • http://pixinsight.com/
Hi all,

Yesterday we released updates for the standard ImageRegistration and Geometry modules with new versions of the StarAlignment, DynamicCrop, Resample and Rotation tools. These new versions introduce an important new feature: Lanczos pixel interpolation.

We are updating the reference documentation for these tools to reflect the changes. The new documents will be distributed through the update system once completed. For now, I have finished a new version of the documentation on pixel interpolation algorithms in PixInsight, where you'll find exhaustive information on important topics that anybody involved in astronomical imaging should know IMO:

http://pixinsight.com/doc/docs/InterpolationAlgorithms/InterpolationAlgorithms.html

Besides a good read (I hope), I have included a number of examples and comparisons at the end of this document that I strongly encourage you to explore. They are intended to provide essential information on critical image processing topics.

Starting with these releases, Lanczos interpolation supersedes bicubic spline on the whole PixInsight platform for all geometric transformations not requiring strong downsampling operations; for the latter tasks, Mitchell-Netravali cubic filter interpolation remains the default option. The advantages of the Lanczos algorithm are of great importance, as shown on the examples included in the above document:

- Higher subpixel accuracy.
- Better preservation of small-scale structures.
- Less generation of aliasing artifacts.

These features are particularly critical to image registration tasks. In all of our tests, Lanczos-3 interpolation systematically leads to a smaller PSF in all images registered with StarAlignment, compared to bicubic spline. For this reason, Lanczos-3 is now the default interpolation algorithm in the StarAlignment tool for image registration tasks not involving strong scale reductions.

This is a new step forward to improve the capabilities of PixInsight to help our users get the most out of their images. We hope you'll like it.
Juan Conejero
PixInsight Development Team
http://pixinsight.com/

Offline Nocturnal

  • PixInsight Jedi Council Member
  • *******
  • Posts: 2727
    • View Profile
    • http://www.carpephoton.com
Congrats on this fine new feature!
Best,

    Sander
---
Edge HD 1100
QHY-8 for imaging, IMG0H mono for guiding, video cameras for occulations
ASI224, QHY5L-IIc
HyperStar3
WO-M110ED+FR-III/TRF-2008
Takahashi EM-400
PIxInsight, DeepSkyStacker, PHD, Nebulosity

Offline RBA

  • PixInsight Guru
  • ****
  • Posts: 511
    • View Profile
    • DeepSkyColors
Second that!

Offline FunTomas

  • PixInsight Addict
  • ***
  • Posts: 135
    • View Profile
    • Astrofoto.sk
Great job Juan. Thank you very much for your effort.

Offline Juan Conejero

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 7036
    • View Profile
    • http://pixinsight.com/
Thank you for your nice comments. Please give a read to this post on the bug reports board:

http://pixinsight.com/forum/index.php?topic=3882.msg26659#msg26659

where interpolation ringing problems have been reported. The balancing between ringing, detail preservation and aliasing in pixel interpolation is a crucial topic to all geometric transformations, especially to image registration. We have addressed this problem with an interpolation clamping feature that can be controlled with the clamping threshold parameter.
Juan Conejero
PixInsight Development Team
http://pixinsight.com/