A new algorithm for the motion vector based interpolation of temporal intermediate images is proposed in this paper. This algorithm uses dedicated weighted median (WM) filters for each interpolation phase (polyphase WM filters) and therefore allows a correction of faulty estimated motion vectors up to a certain degree. Since motion estimation on natural image scenes always suffers from errors in the estimated motion vector field, this ro-bustness of the proposed algorithm against motion vector errors is essential to achieve a high interpolation quality. A new design method for error tolerant polyphase WM filters is presented in the paper. Combining the polyphase WM filters with a spatial nonlinear band separation allows a combination of the error tol-erance of the WM filters with the detail preservation advantages of other interpolation techniques. The results of the proposed algorithm are compared to existing interpolation algorithms.