Application of a variable filter for presampled modulation transfer function analysis with the edge method.

Abstract

We devised a new noise filtering method to reduce the noise in the line spread function (LSF) for presampled modulation transfer function (MTF) analysis with the edge method. A filter was designed to reduce noise effectively using a position-dependent filter controlled by the boundary frequency b for low-pass filtering, which is calculated by 1/2d (d: distance from the LSF center). In this filtering process, strong filters with very low b can be applied to regions distant from the LSF center, and the region near the LSF center can be maintained simultaneously by a correspondingly high b. Presampled MTF accuracies derived by use of the proposed method and an edge spread function (ESF)-fitting method were compared by use of simulated ESFs with and without noise, resembling a computed radiography (CR) and an indirect-type flat panel detector (FPD), respectively. In addition, the edge images of clinical CR, indirect-type FPD, and direct-type FPD systems were examined. For a simulated ESF without noise, the calculated MTFs of the variable filtering method agreed precisely with the true MTFs. The excellent noise-reduction ability of the variable filter was demonstrated for all simulated noisy ESFs and those of three clinical systems. Although the ESF-fitting method provided excellent noise reduction only for the CR-like simulated ESF with noise, its noise elimination performance could not be demonstrated due to the lesser robustness of the fitting.