Fuzzy reasoning-based directional median filter design

A novel fuzzy reasoning-based directional median filter is proposed to remove the random-value impulse noise efficiently. In the proposed filter, the differences between the current pixel and the neighbors aligning with four edge directions are applied with fuzzy reasoning technique such that the current pixel is determined as one of three types of pixels: impulse noise pixel, detailed pixel and noise-free pixel. This paper also proposes a simple approach to determine the edge and line direction in the current sliding window while the current pixel is detected as a noise in the detailed region. With the acquired information of the edge direction, the proposed directional median filter can remove the impulse noise and preserve the detail simultaneously. The simulation results show that the proposed filter outperforms several existing filter schemes for impulse noise removal in an image.