Bilateral filtering of magnetic resonance phase images

Abstract

High-pass filtering is required for the removal of background field inhomogeneities in magnetic resonance phase images. This high-pass filtering smooths across boundaries between areas with large differences in phase. The most prominent boundary is the surface of the brain where areas with large phase values inside the brain are located close to areas outside the brain where the phase is, on average, zero. Cortical areas, which are of great interest in brain MRI, are therefore often degraded by high-pass filtering. Here, we propose the use of the bilateral filter for the high-pass filtering step. The bilateral filter is essentially a Gaussian filter that stops smoothing at boundaries. We show that the bilateral filter improves image quality at the brain's surface, without sacrificing contrast within the brain.