Abstract
A novel correlation-based filter is presented for de-noising functional magnetic resonance imaging (fMRI) data. Temporal correlation-based exponential weights are defined for spatial smoothing of the data, with bias reduction using estimated noise variance. The proposed scheme is tested on simulated and real fMRI data. Finally, the results are compared with conventional filters. The method is found to be effectively suppressing the Rician noise in fMRI data, while improving the SNR.