Noise reduction in magnetic resonance images using IDFT and TERA model reconstruction

The use of local threshold averaging to reduce high-frequency noise without loss of image resolution is examined for reconstruction using inverse discrete Fourier transform (IDFT) and the transient error reconstruction approach (TERA) modeling. Local threshold averaging consists of averaging the data values of the IDFT reconstruction or the model coefficients of the TERA reconstruction when a local m×m average of these coefficients falls below a noise-related value. This threshold averaging avoids the systematic removal of significant high-frequency data components that results from the application of a window. The results show that the threshold TERA-modeled image has a high resolution than the threshold IDFT image, and that both have a higher resolution than a windowed IDFT image for the same level of signal-to-noise improvement