Direction of arrival (DOA) estimation algorithms in the impulsive noise environment are mostly based on fractional low-order statistics (FLOS), which is not only complicated in calculation, but also poor in adaptable to strong impulsive noise. In order to mitigate the impulsive noise, a new direction finding method-global median value filtering method has recently been proposed, which exploits the characteristic that the impulsive noise appears with randomness. However, this algorithm can not detect more sources than the number of sensors. In this paper, a median filtering method based on non-circular signals is proposed, and it is extended to nested array scenarios, which expands the virtual array aperture and improves the number of estimated sources and DOA estimation performance. Then the traditional second order moment DOA estimation method can be used to process the covariance matrix of the filtered received data. To further reduce the computational complexity of this algorithm, the reduced dimensional MUSIC (RD-MUSIC) method is introduced for DOA estimation. Simulation results show that the method has better DOA estimation performance than existing methods, especially in the strong impulsive noise environment.