Abstract：For aquring the mean line of workpiece form

a new kind of digital smoothing filters based on cubic generalized spline function is proposed. Firstly

generalized B spline functions are derived from the linear differential equation. Combining with generalized B spline functions and variational principles

the digital smoothing filters are deduced which compose of IIR filter and indirect generalize B spline transform. Then recursive algorithms of these filters are presented concretely

and the filter’s universal property is testified. Finally

This filter process is appied on a workpiece to acqure mean line. Experimental results indicate these filters take on a better cut-off property than the Gauss fiter

and total calculation only requires 0.002 second for 11200 data points on PC. These filters possess adjustable smoothing property and zero phase transmission

and overcome the drawbacks of low curvature and several oscillations of classical cubic spline interpolation. The recusive algorithm is a efficient and simple.