That’s a fascinating problem. I wonder what a 2D FFT would give you?
I’m not sure I know how to do a 2d fft on these (x,y) pairs, but I can do one for their 2D histogram. What do you think I should be looking for? cheers!
The article by Al-Sharadqah and N. Chernov [2009] Error analysis for circle fitting algorithms, reviews methods that work well for partial datasets (ex: arc of circle). N.Chernov has published code in Matlab for different methods, including Taubin (1991) method. See function: CircleFitByTaubin()
I have implemented Kasa’s and Taubin’s method in the CircleFit.jl package.
Taubin’s method could be optimized though, as it is not using a newton iteration to find the smallest non negative eigenvalue.
Can you share the data for this? I have an idea and it could be fun to do while my computer is crunching way too many numbers on my other cores.
Little late, but geometric algebra can solve this problem using the CGA model.