In some situations, despite the RMS fit error being sub-pixel, if I tweak a point by a pixel or two or add more points it keeps failing to fit saying the image is rotated 180 (see image below). It’s super sensitive and even complains if I move a point to the blue ‘suggested’ spot for the fit.
This seems to be due to limitations in the opencv fitting routines. Writing new fitting routines to overcome this problem would probably be a significant undertaking, but is logged here for the record.
In some situations, despite the RMS fit error being sub-pixel, if I tweak a point by a pixel or two or add more points it keeps failing to fit saying the image is rotated 180 (see image below). It’s super sensitive and even complains if I move a point to the blue ‘suggested’ spot for the fit.
This seems to be due to limitations in the opencv fitting routines. Writing new fitting routines to overcome this problem would probably be a significant undertaking, but is logged here for the record.