Noise is an important factor that degrades the quality of medical images. Impulse noise is a common noise, which is caused by malfunctioning of sensor elements or errors in the transmission of images. In medical images due to presence of white foreground and black background, many pixels have intensities similar to impulse noise and distinction between noisy and regular pixels is difficult. In software techniques, the accuracy of the noise removal is more important than the algorithm's complexity. But for hardware implementation having a low complexity algorithm with an acceptable accuracy is essential. In this paper a low complexity de-noising method is proposed that removes the noise by local analysis of the image blocks. The proposed method distinguishes non-noisy pixels that have noise-like intensities. All steps are designed to have low hardware complexity. Simulation results show that for different magnetic resonance images, the proposed method removes impulse noise with an acceptable accuracy.