Volume of Fluid (VOF) is a widely employed method for multiphase flows prediction for its simplicity, good mass conservation characteristics and natural handling of topologically complex interfaces. For surface tension dominated flows, however, literature has shown that accuracy in surface tension estimations is still an issue, what may cause parasitic currents and inaccurate prediction of pressure jump condition across interfaces. It occurs mainly due to abrupt changes in the volume fraction field across the interfaces, which takes to inaccurate estimates of interfacial curvatures. Therefore, different approaches have been proposed to mitigate this problem including height-functions, volume fraction smoothing, parabolic fittings, among others. This work proposes a novel approach for curvature estimation in VOF, but not limited to it, that sheds a new light on this persistent problem. The idea is to sample the interfaces with clouds of points and normals at the 0.5 level isosurface of the volume fraction field and to compute the curvature for each point of the cloud by a Computer Graphics technique (normal fitting). The curvatures are then projected onto the Eulerian grid in a Front-Tracking fashion. The new method was implemented in the standard OpenFOAM VOF solver (interFoam) resulting in improvements on the pressure jump estimations and in significant reduction of spurious currents. Numerical simulations were performed and results compared to benchmark data showing the feasibility of the idea.