Skip to end of metadata
Go to start of metadata

Page showcasing preliminary work for automated alignment of graphene sheets under an electron microscope. I'm matching the images 64 to 65 (smallest gap), and images 64 to 72 (biggest gap).

Michael Milford | Ben Upcroft | Jamie Warner

Questions / Notes

  • I'm just applying an affine geometric transform once I find feature correspondences, are these images co-planar? Is the image rectified? Affine is probably overkill and suboptimal
  • This is a quick test using low resolution versions of the images, higher accuracy would be very achievable
  • Is the priority alignment using the regular grid, or using the features on the changing feature (I'm guessing the former)?
  • Is the ideal a fully unsupervised or semi-supervised process?
  • When can we publish in Nature Methods? (smile)
  • Do you have more challenging data? These results probably wouldn't impress people that much?

Original unaligned images (Images 64 - 72)

Top left image is 64, top second from left image is 65, bottom image is 72.

Sample Feature Matches using Sequence-based Methods

An affine geometric transform is calculated using these feature correspondences.

64-65

64-72

Auto-aligned and original animated gifs

64-65

Auto-alignedOriginal unalignedManually aligned

64-72

Auto-alignedOriginal unalignedManually aligned

 

 

  • No labels