tilt image center/dimensions for projection_model_to_projection_matrices?

[McHaillet]

Hey @brisvag @alisterburt ,

I had a question about generating matrices from a projectionmodel. I was wondering why the projection_model_to_projection_matrices() function requires a tilt image center but tomogram dimensions. Is there a specific reason for this?
My intuition would be to give tilt image dimensions and tomogram dimensions, or vice versa (both centers).

[alisterburt]

huh I forgot I even added this 😆

no good reason, just a poor design decision! Let's go with dimensions for both

[McHaillet]

Lol!

no good reason, just a poor design decision! Let's go with dimensions for both

Or does center make more sense? 😕 That gives the user more control over the center point, e.g. size / 2, size // 2 + 1

[alisterburt]

Thinking out loud - the shifts in the projection model are shifts which 'center' the tilt image (so that when backprojected everything lines up), everything lining up depends on the convention for the center so allowing a user to change the tilt image center introduces a possibility for error. If we fix the center then the user of the spec is only responsible for getting the shifts right, not the shifts and the tilt image center

Based on that I think dimensions makes more sense... does that make sense?

[McHaillet]

I like your thinking, that is probably the most straightforward usage! Shall I make a PR for it?

[alisterburt]

would love one! appreciate you spending time on this :-)