Support describing object-centric view transformations in OME-Arrow rather than only storing raw images.
This would provide a common abstraction for generating model-ready views, including transforms such as identity, multiscale, fisheye, log-polar, and affine. Where possible, the design should align with OME-Zarr's coordinate transformation model while supporting nonlinear transformations useful for machine learning.
The goal is to store a reproducible transformation recipe instead of requiring multiple transformed copies of the same image data.
Support describing object-centric view transformations in OME-Arrow rather than only storing raw images.
This would provide a common abstraction for generating model-ready views, including transforms such as identity, multiscale, fisheye, log-polar, and affine. Where possible, the design should align with OME-Zarr's coordinate transformation model while supporting nonlinear transformations useful for machine learning.
The goal is to store a reproducible transformation recipe instead of requiring multiple transformed copies of the same image data.