Sensor undistortion using hyperbolic splines in least squares sense

This paper focuses on the removal of static non-linearity from a set of monotone increasing input-output data points, e.g. from a sensor. Mobius transformations are used to fit a spline to the data points to ensure smoothness and analytic invertibility of the interpolant. A new method is proposed to identify the coefficients of the Mobius transformation so as to minimize the twistiness of the resulting interpolant in a least-squares sense. Simulation results are provided to demonstrate the effectiveness of the method. Performance of the method in the presence of observation noise is also investigated