Best practices for image rotation

When I want to rotate an image, Fiji gives me three options for interpolation. See screenshot:

Screen Shot 2020-04-18 at 6.06.09 PM

Other than the usual, “be consistent,” does anyone have advice on choosing which method to use?

Thank you

Hello Nick,

What is the purpose of your image rotation? Image rotations do come at a cost (see example: https://twitter.com/haesleinhuepf/status/1088546103866388481 ).

Rotations in 90 degree increments are generally safe (does not alter original data) if you have a square image.
Non-square rectangles are generally safe to mirror (180 degree rotation) vertically or horizontally. Rotation in ImageJ on non-square rectangles will just result in different image dimensions.

Best,
George

3 Likes

Hi George,

It’s mainly for figure presentation. i.e. rod-shaped cells oriented vertically.

I don’t think I’ve appreciated the cost of rotation at a non-90˚ increment.

I suppose the best way to handle this would be to do all quantifications on the non-rotated image, and only rotate for figure presentation.

Nick

1 Like

I think that’s a good compromise, considering that rod-shaped cells are unlikely to line themselves up neatly! (my former lab had this issue with another tricky long sample, C. elegans ) The video example is showing the effect of successive rotations, so it is exaggerated compared to a single rotation, but it demonstrates that some information is lost with each transformation.

1 Like