The paper proposes a new approach for filtering out impulse noise in digital color images, focusing on medical diagnostics for magnetic resonance imaging (MRI) scans. The conventional filtering methods operate by applying a noise reduction scheme, generally the median filtering approach, for the center pixel of a suitably chosen window that iteratively slides along the entire image. The method proposed in this paper operates by comparing the center pixel with pixels along the four right angle directions, the north, east, west, and south directions, then applying a marginal median filter along the direction that minimizes the aggregate deviation from the center pixel (hence the title NEWS directional filtering). Simulations using GNU Octave 6.4.0 on Corei3 computer with 4 GB RAM show that the proposed method performs comparably with several directional filtering methods in terms of noise reduction and thus validates the proposed approach. Moreover, the method is validated on real MRI scan images.