How to handle fov_estimate for non-square images?

[CelestronPDEV]

Our sensors (and hence our images) are rectangular, not square. But we know the field of view angular dimensions very accurately. When we call tetra3.solve_from_image(), what should we provide for the fov_estimate angle? The smaller of the camera's (width, height) fov angles? The larger of the two fov angles? Or the diagonal (i.e. sort (fov_width^2+fov_height^2)?

[gustavmpettersson]

All fields of view refer to the width of the image, which is typically (but not necessarily) the larger dimension.