Digital image correlation has become an important and effective non-contact optical full-field strain measurement technique. The strain field obtained directly by image correlation algorithm is full of noise. In this letter, we explore a novel way of actively adding small amount of Gaussian random noise to original displacement field, subsequently utilizing the well-known thin-plate spline smoothing (TPSS) technique to smooth the noised displacement field, and finally differentiating smoothed displacement field to get reliable strain field. The resultant method, named as active noise thin-plate spline smoothing (ANTPSS), outperforms the conventional TPSS and spline least-squares approximation. Moreover, ANTPSS successfully smooths the displacement filed obtained from three-point bending experiment of foam block and generates a reliable inhomogeneous strain field.