Segmentation desk · 2026-11-03
Scan-alignment review points to small bending and limits on correction
A contributor discussing thin-plate-spline deformable alignment said hundreds of patches had been inspected manually to assess whether the approach was worthwhile. The reported changes were mostly small amounts of bending rather than strongly localized warping. The same message identifies a separate limitation: flakes or sheets that physically shifted between scans would not be corrected by that approach. Code publication was promised but not established in the downloaded exchange. These observations describe one contributor's inspection, without a published quantitative evaluation here. They help bound expectations for alignment work, but do not show that deformation correction solves every cross-scan discrepancy.